Efficient Signal Processing Acceleration using OpenCL-based FPGA-GPU Hybrid Cooperation for Reconfigurable ECG Diagnosis

Abstract

With the development of Internet of things (IoT), where humans and machines interact, healthcare that measures and diagnoses bio-signals is advancing. The electrocardiogram (ECG) signal has different normal beat characteristics for each person, and it requires long-term data for detecting abnormalities. In this paper, we increased the detection rate of the normal signals by learning the reference signal, which is the standard for diagnosing ECG signals, as individual-specific signals from existing fixed data. In addition, we proposed an OpenCL-based FPGA-GPU hybrid cooperative platform to efficiently diagnose long-term, large-capacity ECG signals.