Highly Sensitive LM@MWCNT Flexible Piezoresistive Sensor for Signal Recovery and Speech Recognition Enabled by CNN

Highly Sensitive LM@MWCNT Flexible Piezoresistive Sensor for Signal Recovery and Speech Recognition Enabled by CNN | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Highly Sensitive LM@MWCNT Flexible Piezoresistive Sensor for Signal Recovery and Speech Recognition Enabled by CNN Haoyu Li, Chenyu Mou, Xiaolin Ran, Yunlong Wang, Shaojiang Wang, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7518325/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 26 Feb, 2026 Read the published version in Microchimica Acta → Version 1 posted 13 You are reading this latest preprint version Abstract Flexible piezoresistive sensors often suffer from limited sensitivity, insufficient functional layer adhesion, and weak interference resistance, thereby restricting their applicability in physiological monitoring. A highly sensitive flexible sensor was developed by encapsulating Liquid Metal droplets (LM) within multi-walled carbon nanotube (MWCNT) to form LM@MWCNT composite droplets, which effectively overcame the aggregation issue of MWCNT and could be uniformly coated onto a polydimethylsiloxane (PDMS) substrate featuring an undulating microstructure. LM droplets were extruded via a mechanical sintering process to infiltrate the gaps, and the subsequent formation of an oxide layer upon oxygen exposure further strengthened the interfacial adhesion between the LM@MWCNT coating and the PDMS substrate. The assembled LM@MWCNT flexible piezoresistive sensor demonstrated a high sensitivity of 14.81 kPa − 1 , exhibiting rapid response and recovery times of 43 ms and 46 ms, respectively, alongside excellent stability during 3000 cycles. By leveraging convolutional neural network (CNN), pulse signals were effectively processed to enhance their stability, while vibration signals generated during throat vocalization were accurately classified, enabling signal recovery after interference (MAE = 0.007) and precise speech recognition with an accuracy of 0.918. The proposed sensor offers significant potential for real-time physiological monitoring and voice-interactive communication. Deep learning Piezoresistive sensors Liquid metals Carbon nanotube Dispersion stability Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction In recent years, significant progress has been made in flexible sensor technology, showcasing vast application prospects in fields, including, smart wearables, healthcare, and industrial automation[ 1 – 5 ]. Flexible sensors can conform to complex and dynamic environments with their unique flexibility, bendability, and adaptability, making them widely applicable for the detection of various physical quantities, including pressure, liquid leakage, and humidity[ 6 – 11 ]. Among various types, flexible piezoresistive sensors have emerged as particularly promising candidates owing to their low power demand, and superior sensitivity[ 12 – 15 ]. To enhance the sensitivity of piezoresistive flexible sensors, constructing microstructures on elastomeric substrates has become a widely adopted strategy. Commonly employed designs include pyramid-shaped and cylindrical microstructures[ 16 , 17 ], which effectively amplify response of the sensor to external stimuli. Carbon-based nanomaterials have attracted considerable attention in the development of flexible sensors due to their outstanding electrical conductivity, mechanical strength, and chemical stability. Multi walled carbon nanotube (MWCNT) have emerged as one of the most promising conductive fillers due to their high aspect ratio, exceptional tensile strength, and excellent electron transport capabilities[ 18 – 21 ]. These characteristics have made MWCNT highly suitable for applications in environmental sensing, biomedical diagnostics, and industrial monitoring[ 22 – 28 ]. However, the practical deployment of MWCNT in flexible sensor systems is frequently hindered by two critical issues: their inherent tendency to aggregate due to strong van der Waals interactions, which lead to poor dispersion stability, and their weak interfacial adhesion to common polymeric substrates, which compromises mechanical integrity and signal reliability under repeated deformation. To address these challenges, material hybridization strategies have been implemented to mitigate the pronounced aggregation of carbon-based nanomaterials arising from strong van der Waals interactions, thereby enhancing their dispersion stability and compatibility with functional matrices[ 29 ]. In this context, Liquid Metal (LM), particularly gallium-based alloys, has attracted growing attention due to its exceptional electrical and thermal conductivity combined with intrinsic fluidity. Unlike conventional rigid fillers, LM readily conforms to complex and dynamically changing geometries, making it highly promising for applications in flexible sensing and soft electronics, such as pressure sensors and electromagnetic shielding films. Upon contact with oxygen, LM rapidly forms a thin oxide layer that stabilizes the droplets. Nevertheless, this oxide shell inevitably compromises the intrinsic electrical conductivity of LM, thereby reducing the overall electronic performance[ 30 , 31 ]. However, LM and MWCNT each offer distinct advantages for flexible electronics, hybrid strategies that exploit their complementary merits remain largely unexplored. In this study, the LM@MWCNT composite was employed as a functional conductive coating on a microstructured PDMS substrate to fabricate the LM@MWCNT flexible piezoresistive sensor. The results demonstrate that MWCNT encapsulates LM droplets via electrostatic repulsion. The resulting LM@MWCNT droplets possess a high positive surface charge, which significantly enhances suspension stability, with no precipitation observed even after 15 days of settling. The LM@MWCNT suspension was uniformly deposited onto the microstructured PDMS substrate to form the conductive layer. The fluidic LM was extruded to fill voids within the conductive network during subsequent mechanical sintering, while rapid oxidation upon air exposure generated a Ga 2 O 3 layer. This oxide layer acts as a strong interfacial adhesive, firmly anchoring the MWCNT conductive network to the PDMS surface and further strengthening the interfacial adhesion between the LM@MWCNT coating and the microstructured PDMS substrate. To further enhance sensing performance, a sandpaper-templating technique was utilized to replicate irregular biomimetic microstructures onto the PDMS substrate, inspired by the undulating surface morphology of the inner bladder wall. These microscale undulating features enhance the number of conductive pathways formed during mechanical deformation, leading to improved pressure sensitivity. Copper wires were subsequently integrated, and the device was assembled to fabricate a highly Sensitive LM@MWCNT flexible piezoresistive sensor. The resulting flexible sensor features high sensitivity, rapid response, and reliable repeatability. The flexible sensor effectively captures subtle physiological signals, such as pulse waves, as well as mechanical vibrations from the throat. By integrating convolutional neural network (CNN) analysis, the LM@MWCNT flexible piezoresistive sensor enables real-time cardiovascular monitoring and robust signal reconstruction. The flexible sensor ensures reliable physiological signal recognition and effective communication. 2. Experimental 2.1 Materials The materials used in this study include Sylgard 184 polydimethylsiloxane (PDMS) and its curing agent (Dow Corning), MWCNT with an average diameter of 10–20 nm, length of 10–30µm, and a density of 2.1 g/cm³ (Tanfeng Tech. Inc.), and LM composed of 76wt% gallium, 20.5wt% indium, and 13.5wt% tin (Huatai Metal Materials Technology Co., Ltd.). Hydrochloric acid was obtained from Codow Chemical Co., and anhydrous ethanol was purchased from a Chinese supplier. All other chemicals used were of analytical grade and used without further purification. 2.2. Preparation of the LM@MWCNT suspension The preparation of the LM@MWCNT suspension involved three main steps. First, 1 g of LM was added to 20 mL of ethanol and mixed with 0.1 mL of HCl, followed by ultrasonication for 50 minutes to obtain an LM suspension with a concentration of 50 mg/ml. Simultaneously, 20 mg of MWCNT were dispersed in 20 mL of ethanol by ultrasonication for 40 minutes to form a 1 mg/mL MWCNT suspension. Next, equal volumes of the two dispersions were mixed and stirred at 800 rpm for 5 hours using a magnetic stirrer. Finally, the resulting mixture was subjected to a second ultrasonication step to obtain a stable LM@MWCNT suspension. 2.3. Fabrication of the LM@MWCNT-based flexible piezoresistive sensor PDMS and its curing agent were mixed at a weight ratio of 10:1, stirred thoroughly, and poured onto sandpaper templates. The mixture was degassed under vacuum for 30 minutes and then cured in an oven at 60°C. After curing, the PDMS was peeled off from the sandpaper to form microstructured templates with mesh sizes of 100, 360, and 600. The LM@MWCNT suspension was uniformly applied to the microstructured surface to create a conductive coating, which was subsequently bonded to the substrate through mechanical sintering. Copper wires were then attached, followed by assembly and encapsulation, to fabricate flexible piezoresistive sensors with three distinct microstructured surfaces. 2.4. Characterization The Thermo Fisher Scientific K-ALPHA X-ray photoelectron spectrometer (XPS), utilizing Al Kα excitation (hv = 1486.8 eV), was employed for elemental analysis and chemical state characterization. Survey scans were performed at 150 eV, while high-resolution scans for Ga 2p were conducted at 50 eV, with calibration using the C 1s peak at 284.8 eV. Zeta potential was measured using the Anton Paar Litesizer 500, which accurately analyzes the charge characteristics of particles in suspension. The Thermo Scientific Helios 5 CX scanning electron microscope (SEM) with focused ion beam (FIB) was used for high-resolution imaging. The electron beam resolution is up to 0.6 nm (STEM/SEM mode), and the ion beam resolution is 2.5 nm at 30 kV. The flexible piezoresistive sensor was subjected to specific pressures using a push-pull tester (SH-2N, SH-100N, China), and the resulting resistance changes were recorded using a digital source meter (Keithley 2450, USA). 3. Results and discussion 3.1 Sensor fabrication The fabrication of the LM@MWCNT-based flexible piezoresistive sensor consists of two main steps: the preparation of the LM@MWCNT suspension and the construction of the microstructured substrate (Fig. 1 a). The preparation of the LM@MWCNT suspension began with ultrasonic treatment of MWCNTs in ethanol to achieve uniform dispersion. LM was subjected to the same ultrasonic protocol to enhance its dispersibility. The pretreated LM was subsequently introduced into the MWCNT suspension, followed by magnetic stirring and a second round of ultrasonication.The mechanical shear fragmented the LM into microscale droplets that were simultaneously encapsulated by MWCNT. The resulting LM@MWCNT suspension exhibited high stability and was subsequently employed in sensor fabrication. Inspired by the microstructure of the bladder inner wall, sandpaper was selected as a templating material to impart surface roughness to the substrate. A PDMS substrate with microstructured topography was fabricated by casting and curing PDMS against abrasive paper. The LM@MWCNT dispersion was then uniformly drop-cast onto the patterned PDMS surface and dried under ambient conditions. A mechanical sintering process was subsequently applied to induce partial extrusion of the encapsulated LM from the MWCNT shells. The released droplets infiltrated interfacial microvoids and conformed tightly to the substrate due to the intrinsic fluidity of LM, forming a continuous conductive network. Finally, conductive copper wires were affixed, and two microstructured PDMS substrates coated with LM@MWCNT coating were laminated face-to-face, completing the construction of the LM@MWCNT flexible piezoresistive sensor(Fig. 1 b). This architecture ensures strong interfacial adhesion, mechanical flexibility, and stable electrical response under repeated deformation. 3.2 Stability analysis of dispersions and coating Figures 2 d–e illustrate the sedimentation behavior of LM@MWCNT and MWCNT dispersions after standing for 15 days. The results demonstrate that the LM@MWCNT droplets exhibit excellent long-term stability, showing no visible sedimentation even after 15 days of settling. In contrast, significant stratification was observed in the MWCNT dispersion, indicating poor stability. The underlying mechanism governing the stability of LM@MWCNT droplets was explored using X-ray photoelectron spectroscopy (XPS). As shown in Fig. 2 a, the full XPS survey spectrum of LM@MWCNT droplets reveals the overall elemental composition. Figure 2 b shows the high-resolution Ga 2p spectra of both LM and LM@MWCNT droplets. Compared with LM particles, the intensity of the Ga 2 O 3 characteristic peak is significantly reduced in the LM@MWCNT sample, indicating that the carboxyl functional groups on the surface of MWCNT coordinate with Ga 3+ ions, thereby effectively suppressing the oxidation of gallium. To further evaluate the surface charge behavior, zeta potentials of MWCNT, LM, and LM@MWCNT droplets were measured (Fig. 2 c). MWCNT particles exhibited a negative zeta potential of − 11.09 mV, while LM particles showed a strongly positive zeta potential of + 43.47 mV. In contrast, LM@MWCNT droplets demonstrated a zeta potential of + 30.43 mV. These results suggest that MWCNT carry negative charges, LM particles possess positive charges, and LM@MWCNT droplets inherit a net positive surface charge due to the highly electropositive nature of the LM component. These findings suggest that the enhanced stability of LM@MWCNT droplets is resulted from the synergistic effects at the LM–MWCNT interface. Following substrate integration, mechanical sintering induces the rupture of LM@MWCNT droplets, resulting in the extrusion of the encapsulated LM. The released droplets can readily penetrate interfacial voids and conform intimately to the microstructured substrate surface due to the excellent fluidity of LM, thereby forming a continuous and firm interface. Adhesion was further evaluated through a tape-peeling test, and the results revealed that LM@MWCNT subjected to mechanical sintering exhibited significantly enhanced interfacial adhesion. In contrast, pristine MWCNT, with inherently low surface energy, interacts with the substrate primarily via weak van der Waals forces, resulting in poor adhesion and a higher propensity for interfacial debonding (Fig. 2 f–g). This strong interfacial adhesion is resulted from the formation of a Ga 2 O 3 oxide layer on the LM surface upon exposure to air[ 31 , 32 ]. Although the formation of the oxide layer slightly compromises the intrinsic conductivity of the LM, it serves as a firm adhesive interface that securely immobilizes the MWCNT conductive network onto the substrate. As shown in Fig. S1 , the resistance of pristine LM and the LM@MWCNT coating was compared before and after three days, revealing that LM@MWCNT demonstrated enhanced electrical stability, as evidenced by the smaller resistance variations compared to pristine LM. This well-integrated MWCNT architecture effectively maintains charge transport pathways, thereby ensuring stable and reliable electrical performance of the sensor under various mechanical deformations. Taken together, these distinctions in adhesion mechanisms provide valuable insights for the rational design of composite interfaces. 3.3 Surface microstructure of the sensor SEM was employed to investigate the surface microstructures of PDMS substrates replicated using sandpapers of different mesh sizes (Fig. 3 a-c). Relatively fewer but larger microstructures were observed per unit area when 100-mesh sandpaper was used as the molding template. In contrast, 600-mesh sandpaper yielded denser and finer microstructures, indicating a significant influence of template fineness on surface morphology. Figures 3 d-e illustrate the changes in interfacial adhesion of the LM@MWCNT coating before and after mechanical sintering. Prior to mechanical sintering, the composite film exhibits relatively weak adhesion, with poor contact between the coating and the PDMS substrate. The LM is partially extruded from the MWCNT shells and firmly adheres to the substrate surface after mechanical sintering, thereby immobilizing the MWCNT network and improving interfacial bonding. Figure 3 f shows the adhesion behavior of the MWCNT coating on the substrate. Figures 3 g–i present higher-magnification SEM images. In Fig. 3 g, LM is clearly encapsulated by the surrounding MWCNT, forming a stable core–shell structure. Following mechanical sintering, LM is extruded from the shell, promoting close adhesion of the MWCNT to the substrate surface as shown in Fig. 3 h. Figure 3 i shows pristine MWCNT coating exhibiting a more dispersed morphology compared with that observed in Fig. 3 h. The PDMS substrate without any coating is shown in Fig. S2, demonstrating that the coating process does not damage the microstructures. 3.4 Pressure sensing performance Sensitivity (S) is a key parameter for evaluating the performance of flexible piezoresistive sensors, defined as \(\:S=\left(\frac{\varDelta\:I}{{I}_{0}}\right)/\varDelta\:P\) , where \(\:{I}_{0}\) is the initial current. \(\:\varDelta\:I\) is the change in current, and \(\:\varDelta\:P\) is the relative change in pressure. To demonstrate the overall performance of the LM@MWCNT flexible piezoresistive sensor, measurements were conducted using a push-pull tester and a digital source meter. The effect of sandpaper with different grit sizes on the pressure sensing performance was investigated (Figs. 4 a-c). The sensor exhibited the highest sensitivity of 12.55 kPa ⁻¹ and 14.81 kPa ⁻¹ when using 100-grit and 360-grit sandpapers, respectively. However, the sensitivity rapidly decreased to 0.483 kPa ⁻¹ (Fig. 4 d) when the grit size increased to 600. This result indicates that as the grit size increases, the PDMS surface microstructure becomes finer, leading to a corresponding increase in sensitivity. However, it affects the elastic deformation capability of the PDMS when the surface microstructure becomes too fine, thereby reducing the sensor's sensitivity. The flexible piezoresistive sensor, manufactured using 360-grit sandpaper, which exhibited the highest sensitivity, was used for testing. The load response is divided into three stages: the maximum sensitivity is observed within the 0–0.016 kPa pressure range (S 1 = 14.81 kPa ⁻¹ ), followed by a lower sensitivity in the 0.016–0.05 kPa range (S 2 = 4.84 kPa ⁻¹ ), and a further reduction in sensitivity beyond 0.05 kPa (S 3 = 0.016 kPa ⁻¹ ). The rapid decline in sensitivity at higher pressures is resulted from the minimal deformation of the microstructure, indicating that the load has saturated. The "Time-ΔI/I0" curve of the LM@MWCNT sensor demonstrates excellent repeatability under cyclic loading at different pressures (Fig. 4 e-f). Furthermore, the sensor continues to maintain clear and stable output signals when the rate is increased from 0.1 mm/s to 0.5 mm/s at a specific pressure (Fig. 4 g). The LM@MWCNT flexible piezoresistive sensor exhibited a response time of 43 ms and a recovery time of 46 ms (Fig. 4 h), demonstrating ability of the sensor to quickly respond to external pressure changes. Compared with recent studies (Fig. 4 i)[ 33 – 39 ], the LM@MWCNT flexible piezoresistive sensor in this work exhibited higher sensitivity. To evaluate the stability of the MXCNT@LM flexible piezoresistive sensor, 3000 cyclic pressure tests were conducted at a specific pressure. The LM@MWCNT sensor maintained stable and clear output waveforms throughout (Fig. 4 j). The combination of high stability and sensitivity indicates the strong potential for practical applications of the LM@MWCNT flexible piezoresistive sensor. 3.5. Pressure sensing mechanism In this work, the total resistance of the LM@MWCNT-based flexible piezoresistive sensor can be expressed as: \(\:{R}_{total}={R}_{lm@mwcnt}+{R}_{copper\:wires}+{R}_{connect}\) , where \(\:{R}_{lm@mwcnt}\) denotes the intrinsic resistance of the LM@MWCNT conductive coating. \(\:{R}_{copper\:wires}\) refers to the resistance of the copper wires, and \(\:{R}_{connect}\) represents the contact resistance at the interface between two LM@MWCNT-coated microstructured substrates. Among these, the resistance variation under applied pressure is primarily attributed to changes in the interfacial contact resistance \(\:{R}_{connect}\) . The resistance can be expressed by the equation: \(\:R=\rho\:\frac{l}{s}\) , where \(\:\rho\:\) denotes the intrinsic resistivity of the conductive material. \(\:l\) represents the length of the conductive path, and \(\:s\) is the effective contact cross-sectional area. The LM@MWCNT-coated microstructured PDMS substrates experience localized deformation under applied pressure, which gradually increases the interfacial contact area between opposing conductive layers. As these microstructured surfaces compress against one another, the enhanced physical contact facilitates the formation of additional conductive pathways, resulting in a marked decrease in total resistance (Fig. 5 a-c). This change in resistance follows a three-phase trend. At low pressures, the initial compression dramatically enlarges the contact area between adjacent microstructures, producing a sharp drop of resistance and high sensitivity. In the moderate-pressure regime, further deformation increases the contact area and increases the number of conductive junctions, albeit at a reduced rate, yielding moderate sensitivity. At high pressures, the microstructures are nearly flattened, the contact area saturates, and resistance changes become negligible, leading to diminished sensitivity. This multi-stage evolution of contact enables the sensor to effectively distinguish between subtle, thereby supporting precise detection. 3.6. Physiological signal monitoring and communication High-noise environments, present critical challenges to human health and communication. Persistent acoustic exposure impairs autonomic regulation, leading to reduced heart rate variability and cardiovascular stress. Simultaneously, intense background noise compromises voice-based communication, hindering the transmission of urgent signals. To address these issues, the LM@MWCNT flexible piezoresistive sensor is employed for real-time physiological monitoring and vibration-based signal transmission. The sensor captures subtle pulse variations and enables the recovery of missing or degraded pulse signals through CNN-assisted analysis when mounted on the wrist. The sensor detects subtle mechanical vibrations produced during vocalization and transduces these into recognizable signals when positioned on the throat. This enables robust communication by decoding vibration patterns. 3.6.1. Real-time physiological signal monitoring and signal recovery The LM@MWCNT flexible piezoresistive sensor is highly effective in monitoring physiological signals such as pulse and wrist flexion. The LM@MWCNT flexible piezoresistive sensor it can precisely detect pulse signals when it is attached to the wrist (Fig. 6 b). The image consistently displays three identifiable peaks in the radial artery, which are determined to be the pulse (P), tidal (T), and diastolic (D) waves[ 40 ]. These features enable the sensor to be used for real-time cardiovascular monitoring, emergency status assessment, and motion tracking. As weak physiological signals, pulse waves exhibit inherent temporal continuity, with each cycle closely correlated to preceding waveforms. However, they are highly susceptible to distortion and loss due to environmental noise, and device imperfections. Consequently, an analytical model not only captures subtle temporal dependencies but also remains robust against incomplete or corrupted waveforms to ensure reliable signal recovery. To address this challenge, a CNN was employed to effectively extract essential temporal features from sequential data and well suit for large-scale physiological signal modeling. The CNN architecture consisted of an input layer, multiple convolutional layers for feature extraction, pooling layers to reduce dimensionality, fully connected layers for feature integration, and a final prediction layer to output the reconstructed signal (Fig. 6 a). A total of 112,500 pulse signal data points were collected through cyclic pressure testing, with 70% used for model training and the remaining 30% reserved as an independent test set to evaluate predictive performance (Fig. S3). The results show that the trained CNN produced strong agreement between predicted and measured signals (R 2 = 0.995, MAE = 0.007), enabling accurate reconstruction of pulse waveforms, thereby minimizing the interference caused by environmental factors and device imperfections (Fig. 6 c). Such performance underscores its suitability for continuous pulse monitoring and the reliable recovery of corrupted or missing signal segments. 3.6.2. Vibration-based speech signal acquisition and recognition The LM@MWCNT flexible piezoresistive sensor demonstrates high sensitivity in collecting sound signals. Vibration signals generated from speech are collected by placing the LM@MWCNT flexible piezoresistive sensor on the user's throat, capturing the vibrational signal features corresponding to different commands. These signals exhibit significant time-series structure, making them suitable for classification using the CNN model to achieve speech recognition. To meet the speech recognition requirements, vibration signals of common commands such as “SOS,” “Stop,” and “Right,” were collected and a CNN recognition model was constructed. The model structure includes convolutional layers, pooling layers, fully connected layers, and a SoftMax output layer. Users were asked to speak everyday commands such as “SOS”, “Right”, “Sleep”, “Stop” and “Home” and corresponding vibration signals were collected (Fig. 6 e-i). the trained model demonstrated high accuracy (Accuracy = 0.918) and was able to accurately recognize common commands after training on over 8000 collected vibration signals (Fig. 6 d), showcasing the speech recognition ability of the LM@MWCNT flexible piezoresistive sensor in high-noise environments. 4. Conclusion The highly sensitive LM@MWCNT flexible piezoresistive sensor was fabricated by integrating a conductive composite coating with a sandpaper-templated microstructured PDMS substrate. This hybrid approach effectively addressed the inherent challenges of LM and MWCNT by simultaneously enhancing dispersion stability, electrical conductivity, and interfacial adhesion. Drawing inspiration from the undulating morphology of biological tissues, a sandpaper-templated microstructure was incorporated to improve pressure sensitivity and facilitate conductive pathway formation. The resulting sensor demonstrated excellent mechanical flexibility, and electrical stability under cyclic deformation, achieving a high sensitivity of 14.81 kPa − 1 with rapid response and recovery times of 43 ms and 46 ms, respectively, and maintaining stable performance over 3000 cycles. The LM@MWCNT flexible piezoresistive sensor reliably detected subtle physiological signals, including pulse waves and throat vibrations. Additionally, the incorporation of CNN-assisted analysis facilitated reliable recovery of lost or degraded signals (MAE = 0.007) and high-accuracy speech recognition (Accuracy = 0.918), thereby ensuring consistent performance in real-time physiological monitoring and voice-interactive communication. This study offers a viable strategy for advancing flexible sensing technologies in wearable human–machine interfaces. Declarations Funding This work was supported by the Fundamental Research Funds for the Central Universities (25CAFUC03026). This work was supported by National College Students Innovation and Entrepreneurship Training Program (Number: S202510624074). Author Contribution H.L. (Haoyu Li) contributed to conceptualization, methodology, software, original draft preparation, review, and editing, and funding acquisition. C.M. (Chenyu Mou), X.R. (Xiaolin Ran), and Y.W. (Yunlong Wang) contributed to investigation and methodology. S.W. (Shaojiang Wang) and H.L. (Huiru Li) contributed to resources and funding acquisition. W.Q. (Wenfeng Qin) contributed to conceptualization, funding acquisition, and review and editing. J.X. (Jiayu Xie) contributed to resources, and review and editing. Data Availability Data is provided within the manuscript or supplementary information files. References Liu C, Yue L, Wei Y et al (2025) A highly sensitive flexible bimodal sensor with uniform pores based on natural polymers for human-computer interaction. Microchim Acta 192:393. https://doi.org/10.1007/s00604-025-07261-3 Tang A, He J, Li A et al (2024) Ti3C2Tx/AgNPs/Ti3C2Tx coated nylon/spandex fabric sensors for human detection with a wide range of application scenarios. J Mater Sci Mater Electron 35:112. https://doi.org/10.1007/s10854-023-11828-2 Bai W, Zhai J, Zhou S et al (2022) Flexible smart wearable co@C@carbon fabric for efficient electromagnetic shielding, thermal therapy, and human movement monitoring. Ind Eng Chem Res 61:11825–11839. https://doi.org/10.1021/acs.iecr.2c02021 Li A, He J, Wang W et al (2022) Self-heating and hydrophobic nanofiber membrane based on Ti3C2TX MXene/ag nanoparticles/thermoplastic polyurethane for electromagnetic interference shielding and sensing performance. Ind Eng Chem Res 61:15249–15260. https://doi.org/10.1021/acs.iecr.2c02886 Zhong Y, Wu L, Gu F et al (2023) Negative pressure-assisted porous structure with gradient dielectrics design for linearity enhancement of flexible capacitance pressure sensor. Colloids Surf -Physicochem Eng Asp 676:132306. https://doi.org/10.1016/j.colsurfa.2023.132306 Pan S, He J, Wang W et al (2025) Piezoresistive MXene@CNTs Nanofiber Membrane Sensors with Micro-Hemispherical Structures via Template-Assisted Electrospinning for Human Health Monitoring. Ind Eng Chem Res 64:3587–3601. https://doi.org/10.1021/acs.iecr.4c04197 Yan J, Chen M, Tan R et al (2024) Flexible multifunctional MXene@Ag nanowires/cotton fabric inspired by transport of nutrients by roots for electromagnetic shielding, infrared stealth, Joule/solar heating and flame retardancy. J Mater Chem A 12:33162–33176. https://doi.org/10.1039/D4TA06712F Chen S, Yan J, Lin C et al (2024) Double-cross-linking strategy for preparing flexible, robust and multifunctional cellulose aerogels. Ind Crops Prod 222:119985. https://doi.org/10.1016/j.indcrop.2024.119985 Yan R, Qin W, Gong G et al (2025) Lightweight composite aramid nanofiber aerogel with multistage pores and layered structure for acoustic and thermal insulation. J Appl Polym Sci 142:e56850. https://doi.org/10.1002/app.56850 Gong G, Chen W, Yan R et al (2025) Flexible MXene/bamboo cellulose fiber/TPU membrane based humidity sensor for non-contact monitoring of human physiological signals. Cellulose 32:1089–1104. https://doi.org/10.1007/s10570-024-06333-6 Peng H, Qin W, Xue Y et al (2025) Synergistic regulation in aramid nanofiber-based composite aerogels for thermo-acoustic insulating applications. Appl Mater Today 43:102671. https://doi.org/10.1016/j.apmt.2025.102671 Chen W, Qin W, Gong G et al (2024) Deep-learning-based prediction of long-term piezoresistive sensing performance of MXene/aramid nanofiber sensors. Adv Eng Mater 26:2401544. https://doi.org/10.1002/adem.202401544 Jiao P, Mueller J, Raney JR et al (2023) Mechanical metamaterials and beyond. Nat Commun 14:6004. https://doi.org/10.1038/s41467-023-41679-8 Luo Y, Abidian MR, Ahn J-H et al (2023) Technology Roadmap for Flexible Sensors. ACS Nano 17:5211–5295. https://doi.org/10.1021/acsnano.2c12606 Yang Y, Liu Y, Yin R (2024) Fiber/Yarn and Textile-Based Piezoresistive Pressure Sensors. Adv Fiber Mater. https://doi.org/10.1007/s42765-024-00479-5 Wang Y, Chen Z, Mei D et al (2022) Highly sensitive and flexible tactile sensor with truncated pyramid-shaped porous graphene/silicone rubber composites for human motion detection. Compos Sci Technol 217:109078. https://doi.org/10.1016/j.compscitech.2021.109078 Resistive pressure sensor based on cylindrical micro structures in periodically ordered electrospun elastic fibers - IOPscience. https://iopscience.iop.org/article/10.1088/1361-665X/aae041 . Accessed 30 Jun 2025 Lim SJ, Lim HS, Joo Y, Jeon D-Y (2020) Impact of MWCNT concentration on the piezo-impedance response of porous MWCNT/PDMS composites. Sens ACTUATORS -Phys 315:112332. https://doi.org/10.1016/j.sna.2020.112332 Hassouneh SS, Yu L, Skov AL, Daugaard AE (2017) Soft and flexible conductive PDMS/MWCNT composites. J Appl Polym Sci 134:44767. https://doi.org/10.1002/app.44767 Li A, Cui C, Wang W et al (2022) Compressible and sensitive aerogels derived from graphene/waste paper for wearable pressure sensor. J Mater Sci Mater Electron 33:4388–4399. https://doi.org/10.1007/s10854-021-07631-6 Liu Q, Zhang Y, Li A et al (2020) Reduced graphene oxide-coated carbonized cotton fabric wearable strain sensors with ultralow detection limit. J Mater Sci-Mater Electron 31:17233–17248. https://doi.org/10.1007/s10854-020-04278-7 Ge Y, Wang L, Hu B et al (2024) Bacterial cellulose-based composite films with liquid metal/graphene synergistic conductive pathways for superior electromagnetic interference shielding and Joule heating performance. Compos Sci Technol 251:110582. https://doi.org/10.1016/j.compscitech.2024.110582 Zhang Y, Ren E, Tang H et al (2020) Carbon nanotubes/acetylene black/ecoflex with corrugated microcracks for enhanced sensitivity for stretchable strain sensors. J Mater Sci Mater Electron 31:14145–14156. https://doi.org/10.1007/s10854-020-03969-5 Huang K, Ning H, Hu N et al (2020) Ultrasensitive MWCNT/PDMS composite strain sensor fabricated by laser ablation process. Compos Sci Technol 192:108105. https://doi.org/10.1016/j.compscitech.2020.108105 Maity D, Rajavel K, Rajendra Kumar RT (2021) MWCNT enabled smart textiles based flexible and wearable sensor for human motion and humidity monitoring. Cellulose 28:2505–2520. https://doi.org/10.1007/s10570-020-03617-5 Zhang Y, Tang H, Li A et al (2020) Extremely stretchable strain sensors with ultra-high sensitivity based on carbon nanotubes and graphene for human motion detection. J Mater Sci Mater Electron 31:12608–12619. https://doi.org/10.1007/s10854-020-03811-y Zhang Y, Ren E, Li A et al (2021) A porous self-healing hydrogel with an island-bridge structure for strain and pressure sensors. J Mater Chem B 9:719–730. https://doi.org/10.1039/D0TB01926G Zhai J, Zhang Y, Cui C et al (2021) Flexible waterborne polyurethane/cellulose nanocrystal composite aerogels by integrating graphene and carbon nanotubes for a highly sensitive pressure sensor. ACS Sustain Chem Eng 9:14029–14039. https://doi.org/10.1021/acssuschemeng.1c03068 Wang T, Kong W-W, Yu W-C et al (2021) A Healable and Mechanically Enhanced Composite with Segregated Conductive Network Structure for High-Efficient Electromagnetic Interference Shielding. Nano-Micro Lett 13. https://doi.org/10.1007/s40820-021-00693-5 (2022) Preparation of liquid metal circuits on flexible polymers by selective laser ablation: Essential mechanism of non-conductivity in ablation part. Appl Surf Sci 605:154746. https://doi.org/10.1016/j.apsusc.2022.154746 Song H, Kim T, Kang S et al (2020) Ga-Based Liquid Metal Micro/Nanoparticles: Recent Advances and Applications. Small 16:1903391. https://doi.org/10.1002/smll.201903391 Kim J-H, Kim S, Dickey MD et al (2024) Interface of gallium-based liquid metals: oxide skin, wetting, and applications. Nanoscale Horiz 9:1099–1119. https://doi.org/10.1039/D4NH00067F Jiang C-S, Lv R-Y, Zou Y-L, Peng H-L (2024) Flexible pressure sensor with wide pressure range based on 3D microporous PDMS/MWCNTs for human motion detection. Microelectron Eng 283:112105. https://doi.org/10.1016/j.mee.2023.112105 Jing Z, Zhang Q, Cheng Y et al (2020) Highly sensitive, reliable and flexible piezoresistive pressure sensors based on graphene-PDMS @ sponge. J Micromechanics Microengineering 30:085012. https://doi.org/10.1088/1361-6439/ab948f Wang L, Peng H, Wang X et al (2016) PDMS/MWCNT-based tactile sensor array with coplanar electrodes for crosstalk suppression. Microsyst Nanoeng 2:1–8. https://doi.org/10.1038/micronano.2016.65 Wang J, Zhang C, Chen D et al (2020) Fabrication of a Sensitive Strain and Pressure Sensor from Gold Nanoparticle-Assembled 3D-Interconnected Graphene Microchannel-Embedded PDMS. ACS Appl Mater Interfaces 12:51854–51863. https://doi.org/10.1021/acsami.0c16152 Kou H, Zhang L, Tan Q et al (2019) Wireless wide-range pressure sensor based on graphene/PDMS sponge for tactile monitoring. Sci Rep 9:3916. https://doi.org/10.1038/s41598-019-40828-8 Su Y, Zhang W, Chen S et al (2021) Piezoresistive Electronic-Skin Sensors Produced With Self-Channeling Laser Microstructured Silicon Molds. IEEE Trans Electron Devices 68:786–792. https://doi.org/10.1109/TED.2020.3045962 A Flexible Pressure Sensor Based on Multiwalled Carbon Nanotubes (2025) / Polydimethylosiloxane Composite for Wearable Electronic-Skin Application | IEEE Journals & Magazine | IEEE Xplore. https://ieeexplore.ieee.org/document/9936604 . Accessed 22 Apr Lu S-Y, Lee C-L, Lin K-Y, Lin Y-H The acute effect of exposure to noise on cardiovascular parameters in young adults Additional Declarations No competing interests reported. Supplementary Files GraphicalAbstracts.jpg SupportingInformation.docx Cite Share Download PDF Status: Published Journal Publication published 26 Feb, 2026 Read the published version in Microchimica Acta → Version 1 posted Editorial decision: Revision requested 30 Oct, 2025 Reviews received at journal 24 Oct, 2025 Reviews received at journal 18 Oct, 2025 Reviewers agreed at journal 14 Oct, 2025 Reviewers agreed at journal 12 Oct, 2025 Reviews received at journal 05 Oct, 2025 Reviewers agreed at journal 20 Sep, 2025 Reviewers agreed at journal 19 Sep, 2025 Reviewers agreed at journal 19 Sep, 2025 Reviewers invited by journal 09 Sep, 2025 Editor assigned by journal 04 Sep, 2025 Submission checks completed at journal 03 Sep, 2025 First submitted to journal 02 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7518325","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":513897785,"identity":"0e8e708f-0ac5-4bbf-9c4d-ac812a1a383c","order_by":0,"name":"Haoyu Li","email":"","orcid":"","institution":"Civil Aviation Flight University of China","correspondingAuthor":false,"prefix":"","firstName":"Haoyu","middleName":"","lastName":"Li","suffix":""},{"id":513897786,"identity":"194bc844-019e-4826-b5d8-fd877992865f","order_by":1,"name":"Chenyu Mou","email":"","orcid":"","institution":"Civil Aviation Flight University of China","correspondingAuthor":false,"prefix":"","firstName":"Chenyu","middleName":"","lastName":"Mou","suffix":""},{"id":513897787,"identity":"ca72b95a-3601-4876-ae81-d596874c772b","order_by":2,"name":"Xiaolin Ran","email":"","orcid":"","institution":"Civil Aviation Flight University of China","correspondingAuthor":false,"prefix":"","firstName":"Xiaolin","middleName":"","lastName":"Ran","suffix":""},{"id":513897788,"identity":"41986356-3cc8-415c-a645-64c31decf152","order_by":3,"name":"Yunlong Wang","email":"","orcid":"","institution":"Civil Aviation Flight University of China","correspondingAuthor":false,"prefix":"","firstName":"Yunlong","middleName":"","lastName":"Wang","suffix":""},{"id":513897789,"identity":"a93ddae2-f0ff-4263-8ef8-94eb729a613a","order_by":4,"name":"Shaojiang Wang","email":"","orcid":"","institution":"Civil Aviation Flight University of China","correspondingAuthor":false,"prefix":"","firstName":"Shaojiang","middleName":"","lastName":"Wang","suffix":""},{"id":513897790,"identity":"435e9e84-a559-4738-b796-75547baf4618","order_by":5,"name":"Huiru Li","email":"","orcid":"","institution":"Civil Aviation Flight University of China","correspondingAuthor":false,"prefix":"","firstName":"Huiru","middleName":"","lastName":"Li","suffix":""},{"id":513897793,"identity":"c53540f3-7aec-4e2e-a883-f831ff947907","order_by":6,"name":"Wenfeng Qin","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAuUlEQVRIiWNgGAWjYFACxgcMjA02PPzsDURrYTYAakmTkew5QJqWwzYGNxyI1MB3I5n5deWO8zwMNxgYP3zMIUKL5I1kNsuzZ27zMM5uYJacuY0ILQa3848ZNrbd5mGWOcDGzEuclmQ2oJZzPGwSCcRrYX7Y2HaAh4doLZL3H7MxNrYl80jwHGwmzi98Zw4zf2xss7O3P9588MNHYrQwHGBgk4CwGBuIUQ/WwvyBSKWjYBSMglEwUgEATQA4WWA+fG8AAAAASUVORK5CYII=","orcid":"","institution":"Civil Aviation Flight University of China","correspondingAuthor":true,"prefix":"","firstName":"Wenfeng","middleName":"","lastName":"Qin","suffix":""},{"id":513897795,"identity":"9c6c6500-def2-4236-ad58-4b1d2707c1db","order_by":7,"name":"Jiayu Xie","email":"","orcid":"","institution":"Civil Aviation Flight University of China","correspondingAuthor":false,"prefix":"","firstName":"Jiayu","middleName":"","lastName":"Xie","suffix":""}],"badges":[],"createdAt":"2025-09-02 13:38:38","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7518325/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7518325/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00604-026-07882-2","type":"published","date":"2026-02-26T15:57:30+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":91458320,"identity":"30fda291-5479-4a0c-bcff-c51383e6eb7f","added_by":"auto","created_at":"2025-09-16 16:41:14","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1169663,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a).\u003c/strong\u003ePreparation of the LM@MWCNT flexible piezoresistive sensor. \u003cstrong\u003e(b).\u003c/strong\u003eSchematic illustration and assembly of the flexible piezoresistive sensors designed based on the morphology of the inner bladder wall.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/9b1391e1134b485e9818c82d.png"},{"id":91458856,"identity":"10a48bfe-65c1-4242-aa7b-d0ad435c0889","added_by":"auto","created_at":"2025-09-16 16:49:14","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":3884006,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a)\u003c/strong\u003eXPS spectrum of the LM@MWCNT samples. \u003cstrong\u003e(b) \u003c/strong\u003eXPS of the Gallium (Ga) element in the LM and LM@MWCNT samples.\u003cstrong\u003e (c)\u003c/strong\u003e Zeta potential of the MWCNT, LM, and LM@MWCNT samples. \u003cstrong\u003e(d)\u003c/strong\u003e Photos of LM@MWCNT suspension after settling for different times.\u003cstrong\u003e (e)\u003c/strong\u003e Photos of MWCNT suspension after settling for different times. \u003cstrong\u003e(f)\u003c/strong\u003e Tape-peeling test of the LM@MWCNT coating. \u003cstrong\u003e(g)\u003c/strong\u003e Tape-peeling test of the MWCNT coating.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/3fa9af8caf38d2c282834ed9.png"},{"id":91458322,"identity":"eb51480b-5910-46e1-ac4f-3fb963f7eb90","added_by":"auto","created_at":"2025-09-16 16:41:14","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":4477441,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a–c)\u003c/strong\u003eSEM images of PDMS surfaces fabricated using sandpapers with different mesh sizes. \u003cstrong\u003e\u0026nbsp;(d)\u003c/strong\u003e SEM image of the interface between the LM@MWCNT coating and the substrate before mechanical sintering. \u003cstrong\u003e(e)\u003c/strong\u003e SEM image of the interface between the LM@MWCNT coating and the substrate after mechanical sintering. \u003cstrong\u003e(f)\u003c/strong\u003e SEM image of the interface between the MWCNT coating and the PDMS substrate. \u003cstrong\u003e(g)\u003c/strong\u003e High-magnification SEM image of LM@MWCNT particles before mechanical sintering. \u003cstrong\u003e(h)\u003c/strong\u003e High-magnification SEM image after mechanical sintering.\u003cstrong\u003e (i) \u003c/strong\u003eHigh-magnification SEM image of MWCNT coating.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/6005e70b407476c50c96d423.png"},{"id":91458854,"identity":"f47f21e2-cd81-4831-9d55-1f24afa37a0a","added_by":"auto","created_at":"2025-09-16 16:49:14","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":759189,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a-c)\u003c/strong\u003e “Pressure-ΔI/I\u003csub\u003e₀\u003c/sub\u003e” curves of the LM@MWCNT flexible piezoresistive sensors fabricated with sandpaper of different mesh sizes. \u003cstrong\u003e(d)\u003c/strong\u003e Comparison of the sensitivity of sensors with different microstructures. \u003cstrong\u003e(e-f)\u003c/strong\u003e Variation curves of ΔI/I\u003csub\u003e₀\u003c/sub\u003e at different pressure frequencies. \u003cstrong\u003e(g)\u003c/strong\u003e Variation curves of ΔI/I\u003csub\u003e₀\u003c/sub\u003e at different pressure frequencies. \u003cstrong\u003e(h)\u003c/strong\u003e Response and recovery times of the LM@MWCNT flexible piezoresistive sensor. \u003cstrong\u003e(i)\u003c/strong\u003e Performance comparison of the LM@MWCNT flexible piezoresistive sensor with recently reported PDMS-based pressure sensors. \u003cstrong\u003e(j)\u003c/strong\u003e “Time-ΔI/I₀” curves after 3000 cyclic pressure tests.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/7ea294719c3429d82a9f0734.png"},{"id":91458855,"identity":"8c08df6b-4a27-4eec-8217-b07d8a608e4c","added_by":"auto","created_at":"2025-09-16 16:49:14","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":985812,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a) \u003c/strong\u003eSchematic illustration of the pressure sensing behavior of the LM@MWCNT-based flexible sensor. \u003cstrong\u003e(b)\u003c/strong\u003e Evolution of interfacial contact area between conductive layers under increasing pressure. \u003cstrong\u003e(c) \u003c/strong\u003eCross-sectional representation of the sensor structure during compressive deformation.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/8ddb45d27472e93c77e63ce1.png"},{"id":91458326,"identity":"bd019866-41c5-4848-a340-73ae98dbf7fa","added_by":"auto","created_at":"2025-09-16 16:41:14","extension":"jpeg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":916062,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a)\u003c/strong\u003e Diagram of CNN architecture. \u003cstrong\u003e(b) \u003c/strong\u003ePulse signal. \u003cstrong\u003e(c)\u003c/strong\u003e Comparison between actual and predicted values. \u003cstrong\u003e(d)\u003c/strong\u003e Confusion matrix of the CNN algorithm for five different commands.\u003cstrong\u003e (e-i)\u003c/strong\u003e Current signals to different commands.\u003c/p\u003e","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/a2f7efe0e1ed8669fec90cf5.jpeg"},{"id":103766386,"identity":"c350c9fb-d917-47cc-a8b4-46625d7bbfba","added_by":"auto","created_at":"2026-03-02 16:14:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":12839744,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/06ac9f50-5909-4f31-a407-a68d3a9e29fc.pdf"},{"id":91458319,"identity":"e9021ae5-4976-49e4-b619-a7e1178ebcc6","added_by":"auto","created_at":"2025-09-16 16:41:14","extension":"jpg","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":133868,"visible":true,"origin":"","legend":"","description":"","filename":"GraphicalAbstracts.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/d1e40f6d3f099993ad82c177.jpg"},{"id":91458337,"identity":"02a94a5a-3c48-4709-bc95-217bfe27c775","added_by":"auto","created_at":"2025-09-16 16:41:14","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":2781895,"visible":true,"origin":"","legend":"","description":"","filename":"SupportingInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-7518325/v1/2258771755959994b9b76dd6.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Highly Sensitive LM@MWCNT Flexible Piezoresistive Sensor for Signal Recovery and Speech Recognition Enabled by CNN","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eIn recent years, significant progress has been made in flexible sensor technology, showcasing vast application prospects in fields, including, smart wearables, healthcare, and industrial automation[\u003cspan additionalcitationids=\"CR2 CR3 CR4\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Flexible sensors can conform to complex and dynamic environments with their unique flexibility, bendability, and adaptability, making them widely applicable for the detection of various physical quantities, including pressure, liquid leakage, and humidity[\u003cspan additionalcitationids=\"CR7 CR8 CR9 CR10\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Among various types, flexible piezoresistive sensors have emerged as particularly promising candidates owing to their low power demand, and superior sensitivity[\u003cspan additionalcitationids=\"CR13 CR14\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. To enhance the sensitivity of piezoresistive flexible sensors, constructing microstructures on elastomeric substrates has become a widely adopted strategy. Commonly employed designs include pyramid-shaped and cylindrical microstructures[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], which effectively amplify response of the sensor to external stimuli.\u003c/p\u003e\u003cp\u003eCarbon-based nanomaterials have attracted considerable attention in the development of flexible sensors due to their outstanding electrical conductivity, mechanical strength, and chemical stability. Multi walled carbon nanotube (MWCNT) have emerged as one of the most promising conductive fillers due to their high aspect ratio, exceptional tensile strength, and excellent electron transport capabilities[\u003cspan additionalcitationids=\"CR19 CR20\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. These characteristics have made MWCNT highly suitable for applications in environmental sensing, biomedical diagnostics, and industrial monitoring[\u003cspan additionalcitationids=\"CR23 CR24 CR25 CR26 CR27\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. However, the practical deployment of MWCNT in flexible sensor systems is frequently hindered by two critical issues: their inherent tendency to aggregate due to strong van der Waals interactions, which lead to poor dispersion stability, and their weak interfacial adhesion to common polymeric substrates, which compromises mechanical integrity and signal reliability under repeated deformation. To address these challenges, material hybridization strategies have been implemented to mitigate the pronounced aggregation of carbon-based nanomaterials arising from strong van der Waals interactions, thereby enhancing their dispersion stability and compatibility with functional matrices[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn this context, Liquid Metal (LM), particularly gallium-based alloys, has attracted growing attention due to its exceptional electrical and thermal conductivity combined with intrinsic fluidity. Unlike conventional rigid fillers, LM readily conforms to complex and dynamically changing geometries, making it highly promising for applications in flexible sensing and soft electronics, such as pressure sensors and electromagnetic shielding films. Upon contact with oxygen, LM rapidly forms a thin oxide layer that stabilizes the droplets. Nevertheless, this oxide shell inevitably compromises the intrinsic electrical conductivity of LM, thereby reducing the overall electronic performance[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. However, LM and MWCNT each offer distinct advantages for flexible electronics, hybrid strategies that exploit their complementary merits remain largely unexplored.\u003c/p\u003e\u003cp\u003eIn this study, the LM@MWCNT composite was employed as a functional conductive coating on a microstructured PDMS substrate to fabricate the LM@MWCNT flexible piezoresistive sensor. The results demonstrate that MWCNT encapsulates LM droplets via electrostatic repulsion. The resulting LM@MWCNT droplets possess a high positive surface charge, which significantly enhances suspension stability, with no precipitation observed even after 15 days of settling. The LM@MWCNT suspension was uniformly deposited onto the microstructured PDMS substrate to form the conductive layer. The fluidic LM was extruded to fill voids within the conductive network during subsequent mechanical sintering, while rapid oxidation upon air exposure generated a Ga\u003csub\u003e2\u003c/sub\u003eO\u003csub\u003e3\u003c/sub\u003e layer. This oxide layer acts as a strong interfacial adhesive, firmly anchoring the MWCNT conductive network to the PDMS surface and further strengthening the interfacial adhesion between the LM@MWCNT coating and the microstructured PDMS substrate. To further enhance sensing performance, a sandpaper-templating technique was utilized to replicate irregular biomimetic microstructures onto the PDMS substrate, inspired by the undulating surface morphology of the inner bladder wall. These microscale undulating features enhance the number of conductive pathways formed during mechanical deformation, leading to improved pressure sensitivity. Copper wires were subsequently integrated, and the device was assembled to fabricate a highly Sensitive LM@MWCNT flexible piezoresistive sensor.\u003c/p\u003e\u003cp\u003eThe resulting flexible sensor features high sensitivity, rapid response, and reliable repeatability. The flexible sensor effectively captures subtle physiological signals, such as pulse waves, as well as mechanical vibrations from the throat. By integrating convolutional neural network (CNN) analysis, the LM@MWCNT flexible piezoresistive sensor enables real-time cardiovascular monitoring and robust signal reconstruction. The flexible sensor ensures reliable physiological signal recognition and effective communication.\u003c/p\u003e"},{"header":"2. Experimental","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Materials\u003c/h2\u003e\u003cp\u003eThe materials used in this study include Sylgard 184 polydimethylsiloxane (PDMS) and its curing agent (Dow Corning), MWCNT with an average diameter of 10\u0026ndash;20 nm, length of 10\u0026ndash;30\u0026micro;m, and a density of 2.1 g/cm\u0026sup3; (Tanfeng Tech. Inc.), and LM composed of 76wt% gallium, 20.5wt% indium, and 13.5wt% tin (Huatai Metal Materials Technology Co., Ltd.). Hydrochloric acid was obtained from Codow Chemical Co., and anhydrous ethanol was purchased from a Chinese supplier. All other chemicals used were of analytical grade and used without further purification.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2. Preparation of the LM@MWCNT suspension\u003c/h2\u003e\u003cp\u003eThe preparation of the LM@MWCNT suspension involved three main steps. First, 1 g of LM was added to 20 mL of ethanol and mixed with 0.1 mL of HCl, followed by ultrasonication for 50 minutes to obtain an LM suspension with a concentration of 50 mg/ml. Simultaneously, 20 mg of MWCNT were dispersed in 20 mL of ethanol by ultrasonication for 40 minutes to form a 1 mg/mL MWCNT suspension. Next, equal volumes of the two dispersions were mixed and stirred at 800 rpm for 5 hours using a magnetic stirrer. Finally, the resulting mixture was subjected to a second ultrasonication step to obtain a stable LM@MWCNT suspension.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3. Fabrication of the LM@MWCNT-based flexible piezoresistive sensor\u003c/h2\u003e\u003cp\u003ePDMS and its curing agent were mixed at a weight ratio of 10:1, stirred thoroughly, and poured onto sandpaper templates. The mixture was degassed under vacuum for 30 minutes and then cured in an oven at 60\u0026deg;C. After curing, the PDMS was peeled off from the sandpaper to form microstructured templates with mesh sizes of 100, 360, and 600. The LM@MWCNT suspension was uniformly applied to the microstructured surface to create a conductive coating, which was subsequently bonded to the substrate through mechanical sintering. Copper wires were then attached, followed by assembly and encapsulation, to fabricate flexible piezoresistive sensors with three distinct microstructured surfaces.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4. Characterization\u003c/h2\u003e\u003cp\u003eThe Thermo Fisher Scientific K-ALPHA X-ray photoelectron spectrometer (XPS), utilizing Al Kα excitation (hv\u0026thinsp;=\u0026thinsp;1486.8 eV), was employed for elemental analysis and chemical state characterization. Survey scans were performed at 150 eV, while high-resolution scans for Ga 2p were conducted at 50 eV, with calibration using the C 1s peak at 284.8 eV. Zeta potential was measured using the Anton Paar Litesizer 500, which accurately analyzes the charge characteristics of particles in suspension. The Thermo Scientific Helios 5 CX scanning electron microscope (SEM) with focused ion beam (FIB) was used for high-resolution imaging. The electron beam resolution is up to 0.6 nm (STEM/SEM mode), and the ion beam resolution is 2.5 nm at 30 kV.\u003c/p\u003e\u003cp\u003eThe flexible piezoresistive sensor was subjected to specific pressures using a push-pull tester (SH-2N, SH-100N, China), and the resulting resistance changes were recorded using a digital source meter (Keithley 2450, USA).\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results and discussion","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Sensor fabrication\u003c/h2\u003e\u003cp\u003eThe fabrication of the LM@MWCNT-based flexible piezoresistive sensor consists of two main steps: the preparation of the LM@MWCNT suspension and the construction of the microstructured substrate (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea).\u003c/p\u003e\u003cp\u003eThe preparation of the LM@MWCNT suspension began with ultrasonic treatment of MWCNTs in ethanol to achieve uniform dispersion. LM was subjected to the same ultrasonic protocol to enhance its dispersibility. The pretreated LM was subsequently introduced into the MWCNT suspension, followed by magnetic stirring and a second round of ultrasonication.The mechanical shear fragmented the LM into microscale droplets that were simultaneously encapsulated by MWCNT. The resulting LM@MWCNT suspension exhibited high stability and was subsequently employed in sensor fabrication. Inspired by the microstructure of the bladder inner wall, sandpaper was selected as a templating material to impart surface roughness to the substrate. A PDMS substrate with microstructured topography was fabricated by casting and curing PDMS against abrasive paper. The LM@MWCNT dispersion was then uniformly drop-cast onto the patterned PDMS surface and dried under ambient conditions. A mechanical sintering process was subsequently applied to induce partial extrusion of the encapsulated LM from the MWCNT shells. The released droplets infiltrated interfacial microvoids and conformed tightly to the substrate due to the intrinsic fluidity of LM, forming a continuous conductive network. Finally, conductive copper wires were affixed, and two microstructured PDMS substrates coated with LM@MWCNT coating were laminated face-to-face, completing the construction of the LM@MWCNT flexible piezoresistive sensor(Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb). This architecture ensures strong interfacial adhesion, mechanical flexibility, and stable electrical response under repeated deformation.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Stability analysis of dispersions and coating\u003c/h2\u003e\u003cp\u003eFigures\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed\u0026ndash;e illustrate the sedimentation behavior of LM@MWCNT and MWCNT dispersions after standing for 15 days. The results demonstrate that the LM@MWCNT droplets exhibit excellent long-term stability, showing no visible sedimentation even after 15 days of settling. In contrast, significant stratification was observed in the MWCNT dispersion, indicating poor stability.\u003c/p\u003e\u003cp\u003eThe underlying mechanism governing the stability of LM@MWCNT droplets was explored using X-ray photoelectron spectroscopy (XPS). As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea, the full XPS survey spectrum of LM@MWCNT droplets reveals the overall elemental composition. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb shows the high-resolution Ga 2p spectra of both LM and LM@MWCNT droplets. Compared with LM particles, the intensity of the Ga\u003csub\u003e2\u003c/sub\u003eO\u003csub\u003e3\u003c/sub\u003e characteristic peak is significantly reduced in the LM@MWCNT sample, indicating that the carboxyl functional groups on the surface of MWCNT coordinate with Ga\u003csup\u003e3+\u003c/sup\u003e ions, thereby effectively suppressing the oxidation of gallium. To further evaluate the surface charge behavior, zeta potentials of MWCNT, LM, and LM@MWCNT droplets were measured (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec). MWCNT particles exhibited a negative zeta potential of \u0026minus;\u0026thinsp;11.09 mV, while LM particles showed a strongly positive zeta potential of +\u0026thinsp;43.47 mV. In contrast, LM@MWCNT droplets demonstrated a zeta potential of +\u0026thinsp;30.43 mV. These results suggest that MWCNT carry negative charges, LM particles possess positive charges, and LM@MWCNT droplets inherit a net positive surface charge due to the highly electropositive nature of the LM component. These findings suggest that the enhanced stability of LM@MWCNT droplets is resulted from the synergistic effects at the LM\u0026ndash;MWCNT interface.\u003c/p\u003e\u003cp\u003eFollowing substrate integration, mechanical sintering induces the rupture of LM@MWCNT droplets, resulting in the extrusion of the encapsulated LM. The released droplets can readily penetrate interfacial voids and conform intimately to the microstructured substrate surface due to the excellent fluidity of LM, thereby forming a continuous and firm interface. Adhesion was further evaluated through a tape-peeling test, and the results revealed that LM@MWCNT subjected to mechanical sintering exhibited significantly enhanced interfacial adhesion. In contrast, pristine MWCNT, with inherently low surface energy, interacts with the substrate primarily via weak van der Waals forces, resulting in poor adhesion and a higher propensity for interfacial debonding (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ef\u0026ndash;g). This strong interfacial adhesion is resulted from the formation of a Ga\u003csub\u003e2\u003c/sub\u003eO\u003csub\u003e3\u003c/sub\u003e oxide layer on the LM surface upon exposure to air[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Although the formation of the oxide layer slightly compromises the intrinsic conductivity of the LM, it serves as a firm adhesive interface that securely immobilizes the MWCNT conductive network onto the substrate. As shown in Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e, the resistance of pristine LM and the LM@MWCNT coating was compared before and after three days, revealing that LM@MWCNT demonstrated enhanced electrical stability, as evidenced by the smaller resistance variations compared to pristine LM. This well-integrated MWCNT architecture effectively maintains charge transport pathways, thereby ensuring stable and reliable electrical performance of the sensor under various mechanical deformations. Taken together, these distinctions in adhesion mechanisms provide valuable insights for the rational design of composite interfaces.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Surface microstructure of the sensor\u003c/h2\u003e\u003cp\u003eSEM was employed to investigate the surface microstructures of PDMS substrates replicated using sandpapers of different mesh sizes (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea-c). Relatively fewer but larger microstructures were observed per unit area when 100-mesh sandpaper was used as the molding template. In contrast, 600-mesh sandpaper yielded denser and finer microstructures, indicating a significant influence of template fineness on surface morphology.\u003c/p\u003e\u003cp\u003eFigures\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed-e illustrate the changes in interfacial adhesion of the LM@MWCNT coating before and after mechanical sintering. Prior to mechanical sintering, the composite film exhibits relatively weak adhesion, with poor contact between the coating and the PDMS substrate. The LM is partially extruded from the MWCNT shells and firmly adheres to the substrate surface after mechanical sintering, thereby immobilizing the MWCNT network and improving interfacial bonding. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ef shows the adhesion behavior of the MWCNT coating on the substrate. Figures\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eg\u0026ndash;i present higher-magnification SEM images. In Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eg, LM is clearly encapsulated by the surrounding MWCNT, forming a stable core\u0026ndash;shell structure. Following mechanical sintering, LM is extruded from the shell, promoting close adhesion of the MWCNT to the substrate surface as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eh. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ei shows pristine MWCNT coating exhibiting a more dispersed morphology compared with that observed in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eh. The PDMS substrate without any coating is shown in Fig. S2, demonstrating that the coating process does not damage the microstructures.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Pressure sensing performance\u003c/h2\u003e\u003cp\u003eSensitivity (S) is a key parameter for evaluating the performance of flexible piezoresistive sensors, defined as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:S=\\left(\\frac{\\varDelta\\:I}{{I}_{0}}\\right)/\\varDelta\\:P\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{I}_{0}\\)\u003c/span\u003e\u003c/span\u003e is the initial current. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varDelta\\:I\\)\u003c/span\u003e\u003c/span\u003e is the change in current, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varDelta\\:P\\)\u003c/span\u003e\u003c/span\u003e is the relative change in pressure. To demonstrate the overall performance of the LM@MWCNT flexible piezoresistive sensor, measurements were conducted using a push-pull tester and a digital source meter. The effect of sandpaper with different grit sizes on the pressure sensing performance was investigated (Figs.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea-c). The sensor exhibited the highest sensitivity of 12.55 kPa\u003csup\u003e⁻\u0026sup1;\u003c/sup\u003e and 14.81 kPa\u003csup\u003e⁻\u0026sup1;\u003c/sup\u003e when using 100-grit and 360-grit sandpapers, respectively. However, the sensitivity rapidly decreased to 0.483 kPa\u003csup\u003e⁻\u0026sup1;\u003c/sup\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed) when the grit size increased to 600. This result indicates that as the grit size increases, the PDMS surface microstructure becomes finer, leading to a corresponding increase in sensitivity. However, it affects the elastic deformation capability of the PDMS when the surface microstructure becomes too fine, thereby reducing the sensor's sensitivity. The flexible piezoresistive sensor, manufactured using 360-grit sandpaper, which exhibited the highest sensitivity, was used for testing. The load response is divided into three stages: the maximum sensitivity is observed within the 0\u0026ndash;0.016 kPa pressure range (S\u003csub\u003e1\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;14.81 kPa\u003csup\u003e⁻\u0026sup1;\u003c/sup\u003e), followed by a lower sensitivity in the 0.016\u0026ndash;0.05 kPa range (S\u003csub\u003e2\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;4.84 kPa\u003csup\u003e⁻\u0026sup1;\u003c/sup\u003e), and a further reduction in sensitivity beyond 0.05 kPa (S\u003csub\u003e3\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.016 kPa\u003csup\u003e⁻\u0026sup1;\u003c/sup\u003e). The rapid decline in sensitivity at higher pressures is resulted from the minimal deformation of the microstructure, indicating that the load has saturated. The \"Time-ΔI/I0\" curve of the LM@MWCNT sensor demonstrates excellent repeatability under cyclic loading at different pressures (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee-f). Furthermore, the sensor continues to maintain clear and stable output signals when the rate is increased from 0.1 mm/s to 0.5 mm/s at a specific pressure (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg). The LM@MWCNT flexible piezoresistive sensor exhibited a response time of 43 ms and a recovery time of 46 ms (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eh), demonstrating ability of the sensor to quickly respond to external pressure changes. Compared with recent studies (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ei)[\u003cspan additionalcitationids=\"CR34 CR35 CR36 CR37 CR38\" citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e], the LM@MWCNT flexible piezoresistive sensor in this work exhibited higher sensitivity. To evaluate the stability of the MXCNT@LM flexible piezoresistive sensor, 3000 cyclic pressure tests were conducted at a specific pressure. The LM@MWCNT sensor maintained stable and clear output waveforms throughout (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ej). The combination of high stability and sensitivity indicates the strong potential for practical applications of the LM@MWCNT flexible piezoresistive sensor.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.5. Pressure sensing mechanism\u003c/h2\u003e\u003cp\u003eIn this work, the total resistance of the LM@MWCNT-based flexible piezoresistive sensor can be expressed as: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{total}={R}_{lm@mwcnt}+{R}_{copper\\:wires}+{R}_{connect}\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{lm@mwcnt}\\)\u003c/span\u003e\u003c/span\u003e denotes the intrinsic resistance of the LM@MWCNT conductive coating. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{copper\\:wires}\\)\u003c/span\u003e\u003c/span\u003e refers to the resistance of the copper wires, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{connect}\\)\u003c/span\u003e\u003c/span\u003e represents the contact resistance at the interface between two LM@MWCNT-coated microstructured substrates. Among these, the resistance variation under applied pressure is primarily attributed to changes in the interfacial contact resistance \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{connect}\\)\u003c/span\u003e\u003c/span\u003e. The resistance can be expressed by the equation: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:R=\\rho\\:\\frac{l}{s}\\)\u003c/span\u003e\u003c/span\u003e, where \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\rho\\:\\)\u003c/span\u003e\u003c/span\u003e denotes the intrinsic resistivity of the conductive material. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:l\\)\u003c/span\u003e\u003c/span\u003e represents the length of the conductive path, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:s\\)\u003c/span\u003e\u003c/span\u003e is the effective contact cross-sectional area.\u003c/p\u003e\u003cp\u003eThe LM@MWCNT-coated microstructured PDMS substrates experience localized deformation under applied pressure, which gradually increases the interfacial contact area between opposing conductive layers. As these microstructured surfaces compress against one another, the enhanced physical contact facilitates the formation of additional conductive pathways, resulting in a marked decrease in total resistance (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea-c). This change in resistance follows a three-phase trend. At low pressures, the initial compression dramatically enlarges the contact area between adjacent microstructures, producing a sharp drop of resistance and high sensitivity. In the moderate-pressure regime, further deformation increases the contact area and increases the number of conductive junctions, albeit at a reduced rate, yielding moderate sensitivity. At high pressures, the microstructures are nearly flattened, the contact area saturates, and resistance changes become negligible, leading to diminished sensitivity. This multi-stage evolution of contact enables the sensor to effectively distinguish between subtle, thereby supporting precise detection.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.6. Physiological signal monitoring and communication\u003c/h2\u003e\u003cp\u003eHigh-noise environments, present critical challenges to human health and communication. Persistent acoustic exposure impairs autonomic regulation, leading to reduced heart rate variability and cardiovascular stress. Simultaneously, intense background noise compromises voice-based communication, hindering the transmission of urgent signals.\u003c/p\u003e\u003cp\u003eTo address these issues, the LM@MWCNT flexible piezoresistive sensor is employed for real-time physiological monitoring and vibration-based signal transmission. The sensor captures subtle pulse variations and enables the recovery of missing or degraded pulse signals through CNN-assisted analysis when mounted on the wrist. The sensor detects subtle mechanical vibrations produced during vocalization and transduces these into recognizable signals when positioned on the throat. This enables robust communication by decoding vibration patterns.\u003c/p\u003e\u003cdiv id=\"Sec14\" class=\"Section3\"\u003e\u003ch2\u003e3.6.1. Real-time physiological signal monitoring and signal recovery\u003c/h2\u003e\u003cp\u003eThe LM@MWCNT flexible piezoresistive sensor is highly effective in monitoring physiological signals such as pulse and wrist flexion. The LM@MWCNT flexible piezoresistive sensor it can precisely detect pulse signals when it is attached to the wrist (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb). The image consistently displays three identifiable peaks in the radial artery, which are determined to be the pulse (P), tidal (T), and diastolic (D) waves[\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. These features enable the sensor to be used for real-time cardiovascular monitoring, emergency status assessment, and motion tracking.\u003c/p\u003e\u003cp\u003eAs weak physiological signals, pulse waves exhibit inherent temporal continuity, with each cycle closely correlated to preceding waveforms. However, they are highly susceptible to distortion and loss due to environmental noise, and device imperfections. Consequently, an analytical model not only captures subtle temporal dependencies but also remains robust against incomplete or corrupted waveforms to ensure reliable signal recovery.\u003c/p\u003e\u003cp\u003eTo address this challenge, a CNN was employed to effectively extract essential temporal features from sequential data and well suit for large-scale physiological signal modeling. The CNN architecture consisted of an input layer, multiple convolutional layers for feature extraction, pooling layers to reduce dimensionality, fully connected layers for feature integration, and a final prediction layer to output the reconstructed signal (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). A total of 112,500 pulse signal data points were collected through cyclic pressure testing, with 70% used for model training and the remaining 30% reserved as an independent test set to evaluate predictive performance (Fig. S3). The results show that the trained CNN produced strong agreement between predicted and measured signals (R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.995, MAE\u0026thinsp;=\u0026thinsp;0.007), enabling accurate reconstruction of pulse waveforms, thereby minimizing the interference caused by environmental factors and device imperfections (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec). Such performance underscores its suitability for continuous pulse monitoring and the reliable recovery of corrupted or missing signal segments.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section3\"\u003e\u003ch2\u003e3.6.2. Vibration-based speech signal acquisition and recognition\u003c/h2\u003e\u003cp\u003eThe LM@MWCNT flexible piezoresistive sensor demonstrates high sensitivity in collecting sound signals. Vibration signals generated from speech are collected by placing the LM@MWCNT flexible piezoresistive sensor on the user's throat, capturing the vibrational signal features corresponding to different commands. These signals exhibit significant time-series structure, making them suitable for classification using the CNN model to achieve speech recognition.\u003c/p\u003e\u003cp\u003eTo meet the speech recognition requirements, vibration signals of common commands such as \u0026ldquo;SOS,\u0026rdquo; \u0026ldquo;Stop,\u0026rdquo; and \u0026ldquo;Right,\u0026rdquo; were collected and a CNN recognition model was constructed. The model structure includes convolutional layers, pooling layers, fully connected layers, and a SoftMax output layer. Users were asked to speak everyday commands such as \u0026ldquo;SOS\u0026rdquo;, \u0026ldquo;Right\u0026rdquo;, \u0026ldquo;Sleep\u0026rdquo;, \u0026ldquo;Stop\u0026rdquo; and \u0026ldquo;Home\u0026rdquo; and corresponding vibration signals were collected (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ee-i). the trained model demonstrated high accuracy (Accuracy\u0026thinsp;=\u0026thinsp;0.918) and was able to accurately recognize common commands after training on over 8000 collected vibration signals (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed), showcasing the speech recognition ability of the LM@MWCNT flexible piezoresistive sensor in high-noise environments.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"4. Conclusion","content":"\u003cp\u003eThe highly sensitive LM@MWCNT flexible piezoresistive sensor was fabricated by integrating a conductive composite coating with a sandpaper-templated microstructured PDMS substrate. This hybrid approach effectively addressed the inherent challenges of LM and MWCNT by simultaneously enhancing dispersion stability, electrical conductivity, and interfacial adhesion. Drawing inspiration from the undulating morphology of biological tissues, a sandpaper-templated microstructure was incorporated to improve pressure sensitivity and facilitate conductive pathway formation. The resulting sensor demonstrated excellent mechanical flexibility, and electrical stability under cyclic deformation, achieving a high sensitivity of 14.81 kPa\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e with rapid response and recovery times of 43 ms and 46 ms, respectively, and maintaining stable performance over 3000 cycles. The LM@MWCNT flexible piezoresistive sensor reliably detected subtle physiological signals, including pulse waves and throat vibrations. Additionally, the incorporation of CNN-assisted analysis facilitated reliable recovery of lost or degraded signals (MAE\u0026thinsp;=\u0026thinsp;0.007) and high-accuracy speech recognition (Accuracy\u0026thinsp;=\u0026thinsp;0.918), thereby ensuring consistent performance in real-time physiological monitoring and voice-interactive communication. This study offers a viable strategy for advancing flexible sensing technologies in wearable human\u0026ndash;machine interfaces.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding\u003c/h2\u003e\u003cp\u003eThis work was supported by the Fundamental Research Funds for the Central Universities (25CAFUC03026).\u003c/p\u003e\u003cp\u003eThis work was supported by National College Students Innovation and Entrepreneurship Training Program (Number: S202510624074).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eH.L. (Haoyu Li) contributed to conceptualization, methodology, software, original draft preparation, review, and editing, and funding acquisition. C.M. (Chenyu Mou), X.R. (Xiaolin Ran), and Y.W. (Yunlong Wang) contributed to investigation and methodology. S.W. (Shaojiang Wang) and H.L. (Huiru Li) contributed to resources and funding acquisition. W.Q. (Wenfeng Qin) contributed to conceptualization, funding acquisition, and review and editing. J.X. (Jiayu Xie) contributed to resources, and review and editing.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData is provided within the manuscript or supplementary information files.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eLiu C, Yue L, Wei Y et al (2025) A highly sensitive flexible bimodal sensor with uniform pores based on natural polymers for human-computer interaction. Microchim Acta 192:393. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00604-025-07261-3\u003c/span\u003e\u003cspan address=\"10.1007/s00604-025-07261-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTang A, He J, Li A et al (2024) Ti3C2Tx/AgNPs/Ti3C2Tx coated nylon/spandex fabric sensors for human detection with a wide range of application scenarios. J Mater Sci Mater Electron 35:112. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10854-023-11828-2\u003c/span\u003e\u003cspan address=\"10.1007/s10854-023-11828-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBai W, Zhai J, Zhou S et al (2022) Flexible smart wearable co@C@carbon fabric for efficient electromagnetic shielding, thermal therapy, and human movement monitoring. Ind Eng Chem Res 61:11825\u0026ndash;11839. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.iecr.2c02021\u003c/span\u003e\u003cspan address=\"10.1021/acs.iecr.2c02021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi A, He J, Wang W et al (2022) Self-heating and hydrophobic nanofiber membrane based on Ti3C2TX MXene/ag nanoparticles/thermoplastic polyurethane for electromagnetic interference shielding and sensing performance. Ind Eng Chem Res 61:15249\u0026ndash;15260. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.iecr.2c02886\u003c/span\u003e\u003cspan address=\"10.1021/acs.iecr.2c02886\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhong Y, Wu L, Gu F et al (2023) Negative pressure-assisted porous structure with gradient dielectrics design for linearity enhancement of flexible capacitance pressure sensor. Colloids Surf -Physicochem Eng Asp 676:132306. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.colsurfa.2023.132306\u003c/span\u003e\u003cspan address=\"10.1016/j.colsurfa.2023.132306\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePan S, He J, Wang W et al (2025) Piezoresistive MXene@CNTs Nanofiber Membrane Sensors with Micro-Hemispherical Structures via Template-Assisted Electrospinning for Human Health Monitoring. Ind Eng Chem Res 64:3587\u0026ndash;3601. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acs.iecr.4c04197\u003c/span\u003e\u003cspan address=\"10.1021/acs.iecr.4c04197\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYan J, Chen M, Tan R et al (2024) Flexible multifunctional MXene@Ag nanowires/cotton fabric inspired by transport of nutrients by roots for electromagnetic shielding, infrared stealth, Joule/solar heating and flame retardancy. J Mater Chem A 12:33162\u0026ndash;33176. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1039/D4TA06712F\u003c/span\u003e\u003cspan address=\"10.1039/D4TA06712F\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen S, Yan J, Lin C et al (2024) Double-cross-linking strategy for preparing flexible, robust and multifunctional cellulose aerogels. Ind Crops Prod 222:119985. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.indcrop.2024.119985\u003c/span\u003e\u003cspan address=\"10.1016/j.indcrop.2024.119985\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYan R, Qin W, Gong G et al (2025) Lightweight composite aramid nanofiber aerogel with multistage pores and layered structure for acoustic and thermal insulation. J Appl Polym Sci 142:e56850. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/app.56850\u003c/span\u003e\u003cspan address=\"10.1002/app.56850\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGong G, Chen W, Yan R et al (2025) Flexible MXene/bamboo cellulose fiber/TPU membrane based humidity sensor for non-contact monitoring of human physiological signals. Cellulose 32:1089\u0026ndash;1104. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10570-024-06333-6\u003c/span\u003e\u003cspan address=\"10.1007/s10570-024-06333-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePeng H, Qin W, Xue Y et al (2025) Synergistic regulation in aramid nanofiber-based composite aerogels for thermo-acoustic insulating applications. Appl Mater Today 43:102671. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apmt.2025.102671\u003c/span\u003e\u003cspan address=\"10.1016/j.apmt.2025.102671\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen W, Qin W, Gong G et al (2024) Deep-learning-based prediction of long-term piezoresistive sensing performance of MXene/aramid nanofiber sensors. Adv Eng Mater 26:2401544. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/adem.202401544\u003c/span\u003e\u003cspan address=\"10.1002/adem.202401544\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJiao P, Mueller J, Raney JR et al (2023) Mechanical metamaterials and beyond. Nat Commun 14:6004. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-023-41679-8\u003c/span\u003e\u003cspan address=\"10.1038/s41467-023-41679-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLuo Y, Abidian MR, Ahn J-H et al (2023) Technology Roadmap for Flexible Sensors. ACS Nano 17:5211\u0026ndash;5295. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acsnano.2c12606\u003c/span\u003e\u003cspan address=\"10.1021/acsnano.2c12606\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang Y, Liu Y, Yin R (2024) Fiber/Yarn and Textile-Based Piezoresistive Pressure Sensors. Adv Fiber Mater. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s42765-024-00479-5\u003c/span\u003e\u003cspan address=\"10.1007/s42765-024-00479-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Y, Chen Z, Mei D et al (2022) Highly sensitive and flexible tactile sensor with truncated pyramid-shaped porous graphene/silicone rubber composites for human motion detection. Compos Sci Technol 217:109078. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compscitech.2021.109078\u003c/span\u003e\u003cspan address=\"10.1016/j.compscitech.2021.109078\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eResistive pressure sensor based on cylindrical micro structures in periodically ordered electrospun elastic fibers - IOPscience. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://iopscience.iop.org/article/10.1088/1361-665X/aae041\u003c/span\u003e\u003cspan address=\"https://iopscience.iop.org/article/10.1088/1361-665X/aae041\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 30 Jun 2025\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLim SJ, Lim HS, Joo Y, Jeon D-Y (2020) Impact of MWCNT concentration on the piezo-impedance response of porous MWCNT/PDMS composites. Sens ACTUATORS -Phys 315:112332. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.sna.2020.112332\u003c/span\u003e\u003cspan address=\"10.1016/j.sna.2020.112332\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHassouneh SS, Yu L, Skov AL, Daugaard AE (2017) Soft and flexible conductive PDMS/MWCNT composites. J Appl Polym Sci 134:44767. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/app.44767\u003c/span\u003e\u003cspan address=\"10.1002/app.44767\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi A, Cui C, Wang W et al (2022) Compressible and sensitive aerogels derived from graphene/waste paper for wearable pressure sensor. J Mater Sci Mater Electron 33:4388\u0026ndash;4399. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10854-021-07631-6\u003c/span\u003e\u003cspan address=\"10.1007/s10854-021-07631-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu Q, Zhang Y, Li A et al (2020) Reduced graphene oxide-coated carbonized cotton fabric wearable strain sensors with ultralow detection limit. J Mater Sci-Mater Electron 31:17233\u0026ndash;17248. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10854-020-04278-7\u003c/span\u003e\u003cspan address=\"10.1007/s10854-020-04278-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGe Y, Wang L, Hu B et al (2024) Bacterial cellulose-based composite films with liquid metal/graphene synergistic conductive pathways for superior electromagnetic interference shielding and Joule heating performance. Compos Sci Technol 251:110582. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compscitech.2024.110582\u003c/span\u003e\u003cspan address=\"10.1016/j.compscitech.2024.110582\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, Ren E, Tang H et al (2020) Carbon nanotubes/acetylene black/ecoflex with corrugated microcracks for enhanced sensitivity for stretchable strain sensors. J Mater Sci Mater Electron 31:14145\u0026ndash;14156. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10854-020-03969-5\u003c/span\u003e\u003cspan address=\"10.1007/s10854-020-03969-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang K, Ning H, Hu N et al (2020) Ultrasensitive MWCNT/PDMS composite strain sensor fabricated by laser ablation process. Compos Sci Technol 192:108105. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.compscitech.2020.108105\u003c/span\u003e\u003cspan address=\"10.1016/j.compscitech.2020.108105\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaity D, Rajavel K, Rajendra Kumar RT (2021) MWCNT enabled smart textiles based flexible and wearable sensor for human motion and humidity monitoring. Cellulose 28:2505\u0026ndash;2520. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10570-020-03617-5\u003c/span\u003e\u003cspan address=\"10.1007/s10570-020-03617-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, Tang H, Li A et al (2020) Extremely stretchable strain sensors with ultra-high sensitivity based on carbon nanotubes and graphene for human motion detection. J Mater Sci Mater Electron 31:12608\u0026ndash;12619. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10854-020-03811-y\u003c/span\u003e\u003cspan address=\"10.1007/s10854-020-03811-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, Ren E, Li A et al (2021) A porous self-healing hydrogel with an island-bridge structure for strain and pressure sensors. J Mater Chem B 9:719\u0026ndash;730. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1039/D0TB01926G\u003c/span\u003e\u003cspan address=\"10.1039/D0TB01926G\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhai J, Zhang Y, Cui C et al (2021) Flexible waterborne polyurethane/cellulose nanocrystal composite aerogels by integrating graphene and carbon nanotubes for a highly sensitive pressure sensor. ACS Sustain Chem Eng 9:14029\u0026ndash;14039. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acssuschemeng.1c03068\u003c/span\u003e\u003cspan address=\"10.1021/acssuschemeng.1c03068\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang T, Kong W-W, Yu W-C et al (2021) A Healable and Mechanically Enhanced Composite with Segregated Conductive Network Structure for High-Efficient Electromagnetic Interference Shielding. Nano-Micro Lett 13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40820-021-00693-5\u003c/span\u003e\u003cspan address=\"10.1007/s40820-021-00693-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e(2022) Preparation of liquid metal circuits on flexible polymers by selective laser ablation: Essential mechanism of non-conductivity in ablation part. Appl Surf Sci 605:154746. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apsusc.2022.154746\u003c/span\u003e\u003cspan address=\"10.1016/j.apsusc.2022.154746\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSong H, Kim T, Kang S et al (2020) Ga-Based Liquid Metal Micro/Nanoparticles: Recent Advances and Applications. Small 16:1903391. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/smll.201903391\u003c/span\u003e\u003cspan address=\"10.1002/smll.201903391\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKim J-H, Kim S, Dickey MD et al (2024) Interface of gallium-based liquid metals: oxide skin, wetting, and applications. Nanoscale Horiz 9:1099\u0026ndash;1119. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1039/D4NH00067F\u003c/span\u003e\u003cspan address=\"10.1039/D4NH00067F\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJiang C-S, Lv R-Y, Zou Y-L, Peng H-L (2024) Flexible pressure sensor with wide pressure range based on 3D microporous PDMS/MWCNTs for human motion detection. Microelectron Eng 283:112105. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.mee.2023.112105\u003c/span\u003e\u003cspan address=\"10.1016/j.mee.2023.112105\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJing Z, Zhang Q, Cheng Y et al (2020) Highly sensitive, reliable and flexible piezoresistive pressure sensors based on graphene-PDMS @ sponge. J Micromechanics Microengineering 30:085012. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1088/1361-6439/ab948f\u003c/span\u003e\u003cspan address=\"10.1088/1361-6439/ab948f\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang L, Peng H, Wang X et al (2016) PDMS/MWCNT-based tactile sensor array with coplanar electrodes for crosstalk suppression. Microsyst Nanoeng 2:1\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/micronano.2016.65\u003c/span\u003e\u003cspan address=\"10.1038/micronano.2016.65\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang J, Zhang C, Chen D et al (2020) Fabrication of a Sensitive Strain and Pressure Sensor from Gold Nanoparticle-Assembled 3D-Interconnected Graphene Microchannel-Embedded PDMS. ACS Appl Mater Interfaces 12:51854\u0026ndash;51863. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/acsami.0c16152\u003c/span\u003e\u003cspan address=\"10.1021/acsami.0c16152\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKou H, Zhang L, Tan Q et al (2019) Wireless wide-range pressure sensor based on graphene/PDMS sponge for tactile monitoring. Sci Rep 9:3916. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41598-019-40828-8\u003c/span\u003e\u003cspan address=\"10.1038/s41598-019-40828-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSu Y, Zhang W, Chen S et al (2021) Piezoresistive Electronic-Skin Sensors Produced With Self-Channeling Laser Microstructured Silicon Molds. IEEE Trans Electron Devices 68:786\u0026ndash;792. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/TED.2020.3045962\u003c/span\u003e\u003cspan address=\"10.1109/TED.2020.3045962\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eA Flexible Pressure Sensor Based on Multiwalled Carbon Nanotubes (2025) / Polydimethylosiloxane Composite for Wearable Electronic-Skin Application | IEEE Journals \u0026amp; Magazine | IEEE Xplore. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ieeexplore.ieee.org/document/9936604\u003c/span\u003e\u003cspan address=\"https://ieeexplore.ieee.org/document/9936604\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 22 Apr\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLu S-Y, Lee C-L, Lin K-Y, Lin Y-H The acute effect of exposure to noise on cardiovascular parameters in young adults\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"microchimica-acta","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"miac","sideBox":"Learn more about [Microchimica Acta](https://link.springer.com/journal/604)","snPcode":"604","submissionUrl":"https://submission.springernature.com/new-submission/604/3","title":"Microchimica Acta","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Deep learning, Piezoresistive sensors, Liquid metals, Carbon nanotube, Dispersion stability","lastPublishedDoi":"10.21203/rs.3.rs-7518325/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7518325/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eFlexible piezoresistive sensors often suffer from limited sensitivity, insufficient functional layer adhesion, and weak interference resistance, thereby restricting their applicability in physiological monitoring. A highly sensitive flexible sensor was developed by encapsulating Liquid Metal droplets (LM) within multi-walled carbon nanotube (MWCNT) to form LM@MWCNT composite droplets, which effectively overcame the aggregation issue of MWCNT and could be uniformly coated onto a polydimethylsiloxane (PDMS) substrate featuring an undulating microstructure. LM droplets were extruded via a mechanical sintering process to infiltrate the gaps, and the subsequent formation of an oxide layer upon oxygen exposure further strengthened the interfacial adhesion between the LM@MWCNT coating and the PDMS substrate. The assembled LM@MWCNT flexible piezoresistive sensor demonstrated a high sensitivity of 14.81 kPa\u003csup\u003e\u0026minus;\u0026thinsp;1\u003c/sup\u003e, exhibiting rapid response and recovery times of 43 ms and 46 ms, respectively, alongside excellent stability during 3000 cycles. By leveraging convolutional neural network (CNN), pulse signals were effectively processed to enhance their stability, while vibration signals generated during throat vocalization were accurately classified, enabling signal recovery after interference (MAE\u0026thinsp;=\u0026thinsp;0.007) and precise speech recognition with an accuracy of 0.918. The proposed sensor offers significant potential for real-time physiological monitoring and voice-interactive communication.\u003c/p\u003e","manuscriptTitle":"Highly Sensitive LM@MWCNT Flexible Piezoresistive Sensor for Signal Recovery and Speech Recognition Enabled by CNN","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-16 16:41:09","doi":"10.21203/rs.3.rs-7518325/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-10-30T14:40:24+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-24T20:16:02+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-18T04:36:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"178480809325089621272012025678653684063","date":"2025-10-14T17:16:37+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"72907435002112856838128500992921961712","date":"2025-10-12T11:59:55+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-10-05T09:27:35+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"192698011506646556661294801734256391932","date":"2025-09-20T04:19:19+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"197700342263900490783152213353749117541","date":"2025-09-19T12:06:14+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"66074583585440874417295752686840338797","date":"2025-09-19T10:23:01+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-09T10:42:57+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-09-04T07:46:26+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-03T23:18:40+00:00","index":"","fulltext":""},{"type":"submitted","content":"Microchimica Acta","date":"2025-09-02T13:27:11+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"microchimica-acta","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"miac","sideBox":"Learn more about [Microchimica Acta](https://link.springer.com/journal/604)","snPcode":"604","submissionUrl":"https://submission.springernature.com/new-submission/604/3","title":"Microchimica Acta","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"7dc537d9-bb25-4c0a-9e92-fd2393c41b93","owner":[],"postedDate":"September 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-03-02T16:10:49+00:00","versionOfRecord":{"articleIdentity":"rs-7518325","link":"https://doi.org/10.1007/s00604-026-07882-2","journal":{"identity":"microchimica-acta","isVorOnly":false,"title":"Microchimica Acta"},"publishedOn":"2026-02-26 15:57:30","publishedOnDateReadable":"February 26th, 2026"},"versionCreatedAt":"2025-09-16 16:41:09","video":"","vorDoi":"10.1007/s00604-026-07882-2","vorDoiUrl":"https://doi.org/10.1007/s00604-026-07882-2","workflowStages":[]},"version":"v1","identity":"rs-7518325","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7518325","identity":"rs-7518325","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}