Real-time Biosignal Recording and Machine-Learning Analysis System

2022 IEEE 4th International Conference on Artificial Intelligence Circuits and Systems (AICAS)(2022)

引用 1|浏览7
暂无评分
摘要
Biosignal recording and processing systems (BRPSs) are in high demand for numerous applications such as brain-machine interfaces, healthcare, and other clinical applications. However, conventional BRPS can only perform simple operations, such as filtering and denoising, but cannot perform robust machine learning-based analyses in real time. This paper proposes an intelligent BRPS that consists of a signal recording front-end for biosignal acquisition, control and visualization hub, and FPGA board for machine learning acceleration. High-speed Ethernet and PCIe interfaces were used to increase the data transmission rate of the system. Moreover, the integrated accelerator in the FPGA is designed in a single-instruction-multiple-data (SIMD) mode to perform complex machine learning operations in parallel to speed up data-processing tasks. The proposed system is validated for various applications, including EEG-based seizure prediction with a convolutional neural network (CNN), EMG-based gesture recognition with a spiking neural network (SNN), and ECG-based arrhythmia detection with a binary neural network (BNN). Experimental results reveal that this system takes 13 ms to process one-second electrophysiological signals at 512 Hz and 32 channels, thus achieving real-time performance. The proposed BRPS is an open-source and expandable system, and different machine-learning approaches can be configured for diverse applications.
更多
查看译文
关键词
Signal Processing System,Biosignal,Graphical User Interface,AI accelerator,FPGA,Neural Networks
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要