Scholarlink

<p>Cardiovascular disorders (CVDs) are the leading cause of death worldwide, reported to the World Heart Federation (WHF). Early detection can therefore stop potentially fatal heart problems and perhaps save lives. Analyzing electrocardiograms (ECGs) can yield useful diagnostic data for identifying different types of heart arrhythmia. Also, innovative machine learning (ML) procedures are established to perceive the heart arrhythmia using ECG signals. Still, the advanced machine learning models impose a burden on wearable devices due to the computational demands. In this research, the advanced Convolutional Neural Network (CNN) classifier named the Ad-MobilNet model is developed to identify multiple arrhythmias using two publicly available datasets. Before the classification, the signals are denoised by the Fast Normalised Least Mean Square (FNLMS) algorithm, which effectively performs the denoising process. Then, Set Partitioning in Hierarchical Trees (SPIHT) with Discrete Wavelet Transform (DWT) architecture is developed to perform the compression process at the sender side. Finally, the compressed signals are decompressed at the remote healthcare server to perform the classification process using the proposed classifier. The proposed model is implemented at the system level using a Field Programmable Gate Array (FPGA) processor and is simulated using MATLAB and Xilinx Verilog coding. Using a variety of performance metrics, the suggested methodology’s performance is determined. On the PTB-XL and CPSC 2018 datasets, it achieved classification accuracy gains of 96.04% and 98.78% respectively.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

FPGA-enabled advanced deep learning accelerator for multi-class ECG signal classification

  • 期刊论文

  • 发表日期：

    2026年2月12日

查看全文
  • V. B. K. L. Aruna,
  • E. Chitra,
  • M. Padmaja

摘要

Cardiovascular disorders (CVDs) are the leading cause of death worldwide, reported to the World Heart Federation (WHF). Early detection can therefore stop potentially fatal heart problems and perhaps save lives. Analyzing electrocardiograms (ECGs) can yield useful diagnostic data for identifying different types of heart arrhythmia. Also, innovative machine learning (ML) procedures are established to perceive the heart arrhythmia using ECG signals. Still, the advanced machine learning models impose a burden on wearable devices due to the computational demands. In this research, the advanced Convolutional Neural Network (CNN) classifier named the Ad-MobilNet model is developed to identify multiple arrhythmias using two publicly available datasets. Before the classification, the signals are denoised by the Fast Normalised Least Mean Square (FNLMS) algorithm, which effectively performs the denoising process. Then, Set Partitioning in Hierarchical Trees (SPIHT) with Discrete Wavelet Transform (DWT) architecture is developed to perform the compression process at the sender side. Finally, the compressed signals are decompressed at the remote healthcare server to perform the classification process using the proposed classifier. The proposed model is implemented at the system level using a Field Programmable Gate Array (FPGA) processor and is simulated using MATLAB and Xilinx Verilog coding. Using a variety of performance metrics, the suggested methodology’s performance is determined. On the PTB-XL and CPSC 2018 datasets, it achieved classification accuracy gains of 96.04% and 98.78% respectively.