Hybrid CNN-LSTM Network for ECG Classification and Its Software-Hardware Co-Design Approach
Song-Nien Tang, Yuan‐Ho Chen, Yu‐Wei Chang, Yuting Chen, Shuo-Hung Chou
Abstract
The hybrid architecture of a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) model has been progressively applied to the time-series data application. This paper developed a hybrid CNN-LSTM network to classify electrocardiography signals. Moreover, we proposed a software (SW)-hardware (HW) co-design approach using a system-on-chip (SoC) field-programmable gate array (FPGA) platform to implement the hybrid CNN-LSTM inference. In our SoC-FPGA design, the CNN model was completed using the SW program while the LSTM model that employs the block circulant weight matrix was realized using the FPGA HW. An experiment was made to achieve 98.63 % ECG detection accuracy within 208.2 ms using the proposed SW-HW co-design SoC-FPGA approach.