Real-Time ECG Signal Extraction And ASIC Design Based On FastICA

Heart disease is a common disease that threatens human health.In recent years,heart-based real-time monitoring of wearable device has developed rapidly with its advantages of portability,low-cost and low power consumption.In the process of wearable device,variety of noise will be introduced,including electrode drift noise,muscle electrical noise and power frequency noise.Filtering the noise to improve the quality of ECG signal is the key factor for the development of wearable heart monitoring device.The technology of digital filter is a common method for ECG extraction.The main idea of this method if to suppress the interference noise and preserve the information of the frequency band of ECG signal through the processing of frequency domain.However,the effect of this method is limited because the frequency of some noise are mixed with ECG signal.Independent component analysis,as a blind source separation method,uses signal correlation as a measure to separate signals,which provides a new idea for ECG signal extraction.This paper explored the application of independent component analysis in the extraction of real-time ECG signal.The research contents and innovations include:1.Researched on real-time ECG signal extraction algorithm based on FastICA.Proposed a real-time FastICA algorithm based on correlation evaluation to solve the problem that independent component analysis algorithm can not extract real-time signal.The algorithm divided mixed signal into signal segments by FastICA algorithm,and then conducted correlation evaluation on each signal segment.The separated signal which was most similar to ECG signal would be modified and combined with previous signal.Furthermore,removing baseline drift before separating processing can have better performance.2.Designed the ECG signal extraction circuit based on FastICA.Focused on the part of calculation of eigenvalue,inverse square root solution,separation and detection of target signals,and the realization of low cost ASIC circuit.For eigenvalue calculation,an implementation method based on element location and Givens rotation is proposed;In order to solve the inverse square root,a method based on fast inverse square root is proposed;And separation and detection crossed implementation is propose to separate and detect target signal.3.Verified the effectiveness of the algorithm and give the realization of the back-end of circuit.The accuracy of the real-time ECG signal extraction algorithm is up to 96.7%,using MIT-BIH database.The ASIC circuit based on FastICA is implemented in the process of SMIC40nm,which can achieve the area of 0.5814mm2 and the power consumption of 50.7 ?W in the premise of ensuring the real time and accuracy.The real-time ECG extraction algorithm based on FastICA and its ASIC implementation proposed in this paper has important reference value and practical significance for the treatment and analysis of ECG signal.