Time-Frequency Collaborative Denoising for Audio-Magnetotelluric Data Using a Wavelet-Based Residual Network

Time-Frequency Collaborative Denoising for Audio-Magnetotelluric Data Using a Wavelet-Based Residual Network | 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 Time-Frequency Collaborative Denoising for Audio-Magnetotelluric Data Using a Wavelet-Based Residual Network Diyang Wang, Yuan Yuan, Qifeng Xiao, Tara P Banjade, Liang Zhang, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7225229/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract The Audio-Magnetotelluric (AMT) method is a key geophysical technique for mineral resource exploration, but anthropogenic electromagnetic interference severely downgrades the data quality. In recent years, neural networks have shown results superior to traditional methods for AMT time-domain denoising. However, existing approaches often overlook deep-seated signal characteristics, leading to suboptimal performance in processing low-frequency data. To address such limitations, we introduce an innovative time-frequency collaborative network—Wavelet-Based Residual Network (WaveResNet). Distinct from conventional single-domain (time/frequency) processing techniques, WaveResNet incorporates a tailored wavelet convolutional architecture that effectively integrates temporal and spectral attributes of AMT signals. By concatenating features from wavelet-decomposed subcomponents and enabling collaborative learning, the network profoundly exploits coupled time-frequency signatures, notably enhancing separation capability for complex anthropogenic noise. Concurrently, the downsampling effect inherent to wavelet decomposition effectively mitigates processing loss in meaningful signals. Furthermore, WaveResNet synchronously models all four electromagnetic field components, fully leveraging inter-channel correlations. The proposed workflow follows a "Detect-and-Denoise" strategy, where only noisy segments are processed, thereby preserving the integrity of low-noise data. Experiments on both synthetic and field data demonstrate the method effectively identifies and suppresses AMT noise, outperforming existing network-based approaches and offering a novel solution for high-fidelity denoising in environments with strong interference. Audio-Magnetotelluric (AMT) Data Processing Neural NetWorks Wavelet Transform Noise Suppression. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Figure 17 Figure 18 Figure 19 Figure 20 Figure 21 Figure 22 Figure 23 Full Text Additional Declarations Tables 1 and 2 are available in the Supplementary Files section. Supplementary Files Table1.xlsx Table 1.Impact of Different Architecture Designs on Network Performance. Table2.xlsx Table 2.Impact of Varying Sample Lengths. graphicalabstract.png Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 23 Sep, 2025 Reviewers invited by journal 25 Aug, 2025 Editor assigned by journal 07 Aug, 2025 First submitted to journal 05 Aug, 2025 Editorial decision: Major Revision 03 Aug, 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-7225229","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":505020089,"identity":"c1df9cc2-455f-47a7-928b-c99264407b90","order_by":0,"name":"Diyang Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyElEQVRIiWNgGAWjYBACfvkDiQ8SKmzk+NkbiNQiOYPhscGHM2nGkj0HiNRicIPxmeTMlsOJBjcSiLVldnOCNG9DWgLDzccbbzDU2EQT1MIvcyzBmHeHTR7j7LRiC4ZjabkNBG1pyElI5j2TVswsnWMmwdhwmLAWgwP5Hw7zth1ObJM8Q6yWGwmJjTOBWnokeIjUAgzbZAZQIEvwAP2SQIxfgDGY/gMUlfbHD2+88aHGhrAWFEdKJJCiHKKFVB2jYBSMglEwMgAAoZlF+HAIzPYAAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0000-7126-0908","institution":"East China University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Diyang","middleName":"","lastName":"Wang","suffix":""},{"id":505020090,"identity":"419f2ec6-484d-4472-b56d-ee37a04c7915","order_by":1,"name":"Yuan Yuan","email":"","orcid":"","institution":"East China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Yuan","middleName":"","lastName":"Yuan","suffix":""},{"id":505020091,"identity":"729f6594-3f3d-491a-852f-ddd3c020d6e0","order_by":2,"name":"Qifeng Xiao","email":"","orcid":"","institution":"East China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Qifeng","middleName":"","lastName":"Xiao","suffix":""},{"id":505020092,"identity":"efde97ff-edf5-4bf1-b55f-d6b3e3e07be4","order_by":3,"name":"Tara P Banjade","email":"","orcid":"","institution":"East China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Tara","middleName":"P","lastName":"Banjade","suffix":""},{"id":505020093,"identity":"9661d028-7307-442f-af4a-cc10fb715dee","order_by":4,"name":"Liang Zhang","email":"","orcid":"","institution":"Guizhou University","correspondingAuthor":false,"prefix":"","firstName":"Liang","middleName":"","lastName":"Zhang","suffix":""},{"id":505020094,"identity":"0bd174c1-671f-4b7e-9f18-b34678082ef7","order_by":5,"name":"Guang Li","email":"","orcid":"","institution":"East China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Guang","middleName":"","lastName":"Li","suffix":""},{"id":505020095,"identity":"0b324c8c-91aa-4a44-aa31-6fdb39d1bc9b","order_by":6,"name":"Jingtian Tang","email":"","orcid":"","institution":"Central South University","correspondingAuthor":false,"prefix":"","firstName":"Jingtian","middleName":"","lastName":"Tang","suffix":""}],"badges":[],"createdAt":"2025-07-27 09:05:41","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7225229/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7225229/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90351524,"identity":"ff51831a-99d2-4381-8eb8-93d1f1e53360","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1993226,"visible":true,"origin":"","legend":"\u003cp\u003eNoise Removal Workflow: Following AMT four-channel electromagnetic data preprocessing, noise segments are identified via the recognition model, then processed by the denoising model to yield denoised data, with denoised segments finally restitched into their original positions to obtain final output.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/6fc8886f618fea4a3fdc82ca.png"},{"id":90351528,"identity":"a662972a-2cf6-470e-866f-387d7fcda936","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":57724,"visible":true,"origin":"","legend":"\u003cp\u003eData Augmentation Example: Performing translation and scaling operations on noisy data as illustrated to enhance sample diversity.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/9425fab400a6c1bf6a0a0209.png"},{"id":90351525,"identity":"a400f2a6-95d6-4dad-85cb-3c8107b74e1f","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":67038,"visible":true,"origin":"","legend":"\u003cp\u003eNoise Library Example: Nine four-channel electromagnetic noise samples were randomly drawn, with highly correlated characteristics observable among the channels.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/38be1103f3cbcc0079e84935.png"},{"id":90352859,"identity":"71f69511-a764-404b-82f5-d185b70b7243","added_by":"auto","created_at":"2025-09-01 18:28:28","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":732680,"visible":true,"origin":"","legend":"\u003cp\u003eDataset Construction Workflow: Extract 512-point data segments from curated near-clean datasets or synthetic data as labels, while capturing equal-length segments from noise data. The Sample is then formed by additive combination, creating labeled pairs with corresponding labels.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/4f3cabe3c47bfd0d49e3ec58.png"},{"id":90351744,"identity":"10feef69-30a1-4840-a47d-17f3e7755901","added_by":"auto","created_at":"2025-09-01 18:04:28","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":153110,"visible":true,"origin":"","legend":"\u003cp\u003eBasic Network Block Architecture (Block-A): For input signal features, wavelet transform is first applied to obtain multi-scale components. These components are concatenated and processed through a series of convolutional and activation operations for feature extraction. An SE (Squeeze-and-Excitation) module then performs channel-wise weighting, with final residual connection applied to produce the output.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/9a1554a3b426e5c22ec6d64f.png"},{"id":90351746,"identity":"2fea8ebe-75da-49f2-8ac6-86defca98ab0","added_by":"auto","created_at":"2025-09-01 18:04:28","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":32944,"visible":true,"origin":"","legend":"\u003cp\u003eNetwork Basic Structure Block (Block-B): For input signal features, two convolutional operations are sequentially performed for feature extraction, then a residual connection is applied with to produce the output.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/33af2a6a95f43537b0fc2bdf.png"},{"id":90351536,"identity":"ad3311a0-ae98-4760-8df9-a15d3beaf124","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":40189,"visible":true,"origin":"","legend":"\u003cp\u003eOverall Network Architecture: Multiple instances of Block-A and Block-B are modularly stacked, culminating in a fully-connected layer for final output.\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/3c1a74b968a5b8ca60d441f1.png"},{"id":90351538,"identity":"b086c80b-5667-4272-b937-23e9eef133f6","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":622365,"visible":true,"origin":"","legend":"\u003cp\u003eData Processing Pipeline: Input AMT four-channel electromagnetic data is denoised via the proposed method, followed by Fourier transformation to obtain spectral information, ultimately leading to impedance estimation outputs.\u003c/p\u003e","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/c6c139513961a7af9d7fdd60.png"},{"id":90352860,"identity":"9286bde6-3da0-427d-816c-7d4b6fe8590e","added_by":"auto","created_at":"2025-09-01 18:28:28","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":108779,"visible":true,"origin":"","legend":"\u003cp\u003eSimulated Noisy Data: With a duration of 3500 seconds and sampling rate of 150 Hz, severe interference is observed persisting throughout the entire period in the electric channels, while magnetic channels exhibit spiky interference concentrated at specific points\u003c/p\u003e","description":"","filename":"Figure9.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/63401c1b9df8151243a87fe0.png"},{"id":90352395,"identity":"c21c212a-cbd5-4c53-9aab-318b8ce1b69c","added_by":"auto","created_at":"2025-09-01 18:20:28","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":106301,"visible":true,"origin":"","legend":"\u003cp\u003eDenoised Simulated Data Observations: The four-channel electromagnetic data exhibit virtually noise-free characteristics across all channels.\u003c/p\u003e","description":"","filename":"Figure10.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/186ea6c9c4706b8b23a4dd7b.png"},{"id":90351565,"identity":"8b386774-7005-46f8-883a-85afcefe3ffe","added_by":"auto","created_at":"2025-09-01 17:56:29","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":99651,"visible":true,"origin":"","legend":"\u003cp\u003eSegment of Simulated Noisy Data: The electric channels exhibit readily discernible noise contamination, with spatially and temporally correlated anomalous fluctuations observed in the corresponding magnetic channels.\u003c/p\u003e","description":"","filename":"Figure11.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/53ded52253aabd5d95c4d1a4.png"},{"id":90351558,"identity":"ab05fc4c-486a-4d7b-af30-5892a59b796a","added_by":"auto","created_at":"2025-09-01 17:56:29","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":101932,"visible":true,"origin":"","legend":"\u003cp\u003eNoise Identification in Simulated Data Segment: The noise-contaminated intervals, color-coded in red, substantially outline the anomalous regions with near-complete coverage.\u003c/p\u003e","description":"","filename":"Figure12.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/fe9e617bac040d12f9bb7479.png"},{"id":90351557,"identity":"6c9ac737-94cc-4740-adad-239d87461892","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":89597,"visible":true,"origin":"","legend":"\u003cp\u003eResNet Denoising Results: The left panel displays denoised outcomes, while the right panel shows residuals between denoised data and original noise-free data. It is evident that noise in both electric and magnetic channels has been substantially suppressed, with residual magnitudes significantly smaller than the signal amplitudes.\u003c/p\u003e","description":"","filename":"Figure13.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/87793fc2eb0cce237ecd3b3b.png"},{"id":90351543,"identity":"984a3187-08fe-4047-856b-4b0c05091ddf","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":14,"title":"Figure 14","display":"","copyAsset":false,"role":"figure","size":89012,"visible":true,"origin":"","legend":"\u003cp\u003eWaveResNet Denoising Results: The left panel presents denoised outputs, while the right panel displays residual differences between denoised data and pristine noise-free data. Evidently, noise contamination in both electric and magnetic channels has been markedly suppressed, with residual magnitudes significantly lower than those generated by the ResNet benchmark.\u003c/p\u003e","description":"","filename":"Figure14.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/8c16f680ddae292ea7da13b8.png"},{"id":90352207,"identity":"773d45db-f5f6-4bd4-acf2-54918105c2c9","added_by":"auto","created_at":"2025-09-01 18:12:28","extension":"png","order_by":15,"title":"Figure 15","display":"","copyAsset":false,"role":"figure","size":58911,"visible":true,"origin":"","legend":"\u003cp\u003eApparent Resistivity Comparison of Simulated Noisy Data Before/After Denoising: Pre-denoising measurements exhibit discrepant scatter with values markedly exceeding baseline levels. Post-processing through dual networks restores apparent resistivity and phase components to theoretical ranges, though residual oscillations persist at initial frequency points whereas the proposed method demonstrates comparatively enhanced data stability.\u003c/p\u003e","description":"","filename":"Figure15.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/571e4a480c2916c57cca9993.png"},{"id":90352209,"identity":"ae523faa-2a59-4136-aef0-97d541e8249b","added_by":"auto","created_at":"2025-09-01 18:12:28","extension":"png","order_by":16,"title":"Figure 16","display":"","copyAsset":false,"role":"figure","size":100397,"visible":true,"origin":"","legend":"\u003cp\u003eField Measurement Noisy Data: Comprising 3500 seconds of acquisition at 150 Hz sampling rate, gross amplitude contamination persists across electromagnetic channels, exhibiting pronounced inter-channel coherence.\u003c/p\u003e","description":"","filename":"Figure16.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/ca123f672c6a0fb548ba3642.png"},{"id":90351758,"identity":"935516e1-473b-45d5-a1a6-2fc8fe6fa89f","added_by":"auto","created_at":"2025-09-01 18:04:28","extension":"png","order_by":17,"title":"Figure 17","display":"","copyAsset":false,"role":"figure","size":99176,"visible":true,"origin":"","legend":"\u003cp\u003eField Data Post-Denoising: Electromagnetic interference across all four channels has been significantly mitigated, yielding substantially improved baseline stability and signal integrity.\u003c/p\u003e","description":"","filename":"Figure17.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/a60b4b750cfb6f6efb9d82c4.png"},{"id":90351553,"identity":"9599de6c-9515-40a4-b1bd-49ffa3c02eac","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":18,"title":"Figure 18","display":"","copyAsset":false,"role":"figure","size":106914,"visible":true,"origin":"","legend":"\u003cp\u003eField Data Segment: A 60-second segment extracted from the 3500s duration reveals intricate patterns of four-channel electromagnetic interference exhibiting tightly synchronized waveforms and robust phase-synchronization.\u003c/p\u003e","description":"","filename":"Figure18.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/dd46dbce409b86f304ddd948.png"},{"id":90351764,"identity":"6fbf561d-abf6-4176-9542-6b77aa767a54","added_by":"auto","created_at":"2025-09-01 18:04:29","extension":"png","order_by":19,"title":"Figure 19","display":"","copyAsset":false,"role":"figure","size":108399,"visible":true,"origin":"","legend":"\u003cp\u003eNoise Identification in Field Data Segment: Portions color-coded red denote algorithmically recognized noise intervals, superimposed atop the original field data traces rendered in black, demonstrating precise localization and near-complete enclosure of the anomalous regions.\u003c/p\u003e","description":"","filename":"Figure19.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/74d6d35b8158a1399b7b2f4f.png"},{"id":90351552,"identity":"1d2e8474-50a7-4fd0-ae45-e6fdfdd42da7","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":20,"title":"Figure 20","display":"","copyAsset":false,"role":"figure","size":112477,"visible":true,"origin":"","legend":"\u003cp\u003eResNet Denoising Performance: While noise components are effectively suppressed, close examination reveals significantly attenuated amplitudes within denoised regions relative to ambient signal levels, particularly conspicuous in the 45-60 second interval.\u003c/p\u003e","description":"","filename":"Figure20.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/975801a2a773c4bdc04117b2.png"},{"id":90351570,"identity":"556e3970-9561-496a-8111-5df03a628176","added_by":"auto","created_at":"2025-09-01 17:56:29","extension":"png","order_by":21,"title":"Figure 21","display":"","copyAsset":false,"role":"figure","size":112807,"visible":true,"origin":"","legend":"\u003cp\u003eWaveResNet Denoising Performance: Noise contamination has been comprehensively suppressed below instrumental detection thresholds, with amplitude integrity impeccably preserved throughout formerly noise-contaminated segments.\u003c/p\u003e","description":"","filename":"Figure21.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/13d810328820fe25d27b135a.png"},{"id":90351759,"identity":"6cef9576-3b41-49b9-9339-f62e0e719d81","added_by":"auto","created_at":"2025-09-01 18:04:28","extension":"png","order_by":22,"title":"Figure 22","display":"","copyAsset":false,"role":"figure","size":69374,"visible":true,"origin":"","legend":"\u003cp\u003eApparent Resistivity Comparison at Field Site A: Pre-processing data exhibits an asymptotic 45-degree ascent in apparent resistivity tails with phase components approaching zero, indicative of severe near-source contamination. Post remote-reference processing significantly postpones near-source distortion to lower frequencies in both apparent resistivity and phase spectra. Comparative analysis reveals that while both ResNet and the proposed framework extend effective bandwidth towards lower frequencies, the latter achieves deeper penetration into the low-frequency domain with superior proximity to remote-reference benchmarks.\u003c/p\u003e","description":"","filename":"Figure22.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/558832c75c5d4b9835192694.png"},{"id":90351556,"identity":"11f3aaaf-3f41-4a19-8f6a-ea7c9e1d064e","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":23,"title":"Figure 23","display":"","copyAsset":false,"role":"figure","size":72126,"visible":true,"origin":"","legend":"\u003cp\u003ePre-processing profiles manifest asymptotic 45-degree ascent trajectories in resistivity tails accompanied by anomalous phase drifting , indicative of severe anthropogenic interference signatures with fragmented dispersion patterns. Post remote-reference processing restores coherent structure to both resistivity and phase spectra, exhibiting stabilized impedances. Comparison of Post-Processing Results: Both ResNet and the proposed method significantly improve curve morphology; however, the proposed approach yields markedly concentrated apparent resistivity distributions devoid of dispersion anomalies, while demonstrating closer alignment with remote-reference benchmarks.\u003c/p\u003e","description":"","filename":"Figure23.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/6effc0373eb052f6227fe824.png"},{"id":91149284,"identity":"40ba664b-4a49-4251-8e4a-06e0bbfe19ae","added_by":"auto","created_at":"2025-09-12 06:48:30","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3758373,"visible":true,"origin":"","legend":"","description":"","filename":"renamed9bf67.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1_covered_339b6efe-13a5-4f26-a4c9-06824675c502.pdf"},{"id":90351555,"identity":"538dd3e3-7357-4d7a-a7cd-c3600ea8e481","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":10151,"visible":true,"origin":"","legend":"\u003cp\u003eTable 1.Impact of Different Architecture Designs on Network Performance.\u003c/p\u003e","description":"","filename":"Table1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/bbd66fe8e7d9bf840c7d1de1.xlsx"},{"id":90351535,"identity":"7a3f2319-c610-4a56-bc01-3c61ab7539c8","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":10071,"visible":true,"origin":"","legend":"\u003cp\u003eTable 2.Impact of Varying Sample Lengths.\u003c/p\u003e","description":"","filename":"Table2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/c230734d0c6395fb4a4fc2e7.xlsx"},{"id":90351545,"identity":"71a0b782-f040-4d19-bbed-5914479f179e","added_by":"auto","created_at":"2025-09-01 17:56:28","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":619181,"visible":true,"origin":"","legend":"","description":"","filename":"graphicalabstract.png","url":"https://assets-eu.researchsquare.com/files/rs-7225229/v1/35df840ff2fabc2560a826ef.png"}],"financialInterests":"\u003cp\u003eTables 1 and 2 are available in the Supplementary Files section.\u003c/p\u003e","formattedTitle":"\u003cp\u003e\u003cstrong\u003eTime-Frequency Collaborative Denoising for Audio-Magnetotelluric Data Using a Wavelet-Based Residual Network\u003c/strong\u003e\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"earth-planets-and-space","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"epsp","sideBox":"Learn more about [Earth, Planets and Space](http://earth-planets-space.springeropen.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/epsp/default.aspx","title":"Earth, Planets and Space","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Audio-Magnetotelluric (AMT), Data Processing, Neural NetWorks, Wavelet Transform, Noise Suppression.","lastPublishedDoi":"10.21203/rs.3.rs-7225229/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7225229/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe Audio-Magnetotelluric (AMT) method is a key geophysical technique for mineral resource exploration, but anthropogenic electromagnetic interference severely downgrades the data quality. In recent years, neural networks have shown results superior to traditional methods for AMT time-domain denoising. However, existing approaches often overlook deep-seated signal characteristics, leading to suboptimal performance in processing low-frequency data. To address such limitations, we introduce an innovative time-frequency collaborative network\u0026mdash;Wavelet-Based Residual Network (WaveResNet). Distinct from conventional single-domain (time/frequency) processing techniques, WaveResNet incorporates a tailored wavelet convolutional architecture that effectively integrates temporal and spectral attributes of AMT signals. By concatenating features from wavelet-decomposed subcomponents and enabling collaborative learning, the network profoundly exploits coupled time-frequency signatures, notably enhancing separation capability for complex anthropogenic noise. Concurrently, the downsampling effect inherent to wavelet decomposition effectively mitigates processing loss in meaningful signals. Furthermore, WaveResNet synchronously models all four electromagnetic field components, fully leveraging inter-channel correlations. The proposed workflow follows a \"Detect-and-Denoise\" strategy, where only noisy segments are processed, thereby preserving the integrity of low-noise data. Experiments on both synthetic and field data demonstrate the method effectively identifies and suppresses AMT noise, outperforming existing network-based approaches and offering a novel solution for high-fidelity denoising in environments with strong interference.\u003c/p\u003e","manuscriptTitle":"Time-Frequency Collaborative Denoising for Audio-Magnetotelluric Data Using a Wavelet-Based Residual Network","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-01 17:56:23","doi":"10.21203/rs.3.rs-7225229/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2025-09-23T09:12:45+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-25T05:02:49+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-07T17:23:27+00:00","index":"","fulltext":""},{"type":"submitted","content":"Earth, Planets and Space","date":"2025-08-06T00:43:24+00:00","index":"","fulltext":""},{"type":"decision","content":"Major Revision","date":"2025-08-03T08:43:02+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"earth-planets-and-space","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"epsp","sideBox":"Learn more about [Earth, Planets and Space](http://earth-planets-space.springeropen.com)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/epsp/default.aspx","title":"Earth, Planets and Space","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4cd96393-5738-46b4-981f-0cbb81941996","owner":[],"postedDate":"September 1st, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-03-09T03:59:19+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-01 17:56:23","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7225229","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7225229","identity":"rs-7225229","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}