Fpga Based Hardware Entertained Light Weight Self Attention Deep Learning Framework for Better Classification of Ecg Signals

Abstract High Speed Extreme learning networks (HSELN) achieve the two-fold challenges such as (i) reducing the complexity in deep learning algorithms without comprising the diagnosis performance and (ii) sustaining the performance of wearable devices suitable for the continuous monitoring of ECG signals. In order to improve resource constraint factors like lower latency, lower utilization, and energy efficiency, the article also focuses on the deployment of deep learning algorithms utilizing Hardware-Software Co-design techniques. Utilizing a variety of ECG datasets for the extended testing and validating it with the ten-cross validation method. Additionally, the suggested method is implemented in multiple Zynq-SoC families, and variables like utilization, latency, and power are computed and compared to numerous current hardware-centric deep learning architectures. Experimentation demonstrates that the proposed deep learning architecture has shown its excellence in diagnosing the ECG signals and proves to be avital solution for deploying in wearable devices.