Research On QRS Wave Detection Algorithm For Multi-lead Mobile ECG Equipment

According to WHO health data,heart disease has become one of the three major diseases that endanger human health.Since the nineteenth century ECG was invented,ECG automatic analysis technology has been a hot research.In recent years,the popularity of multi-lead mobile ECG equipment is to improve the quality of ECG monitoring a step,but also to the researchers how to further improve the QRS wave detection rate of the problem.The research work of this paper focuses on the QRS wave detection algorithm which is suitable for multi-lead mobile ECG equipment,which mainly includes the following aspects:First of all,this paper analyzes the noise in the ECG signal,including its signal source,type and spectrum range.According to the noise characteristics,we designed a mean filter and an improved median filter to reduce baseline drift and high frequency noise in ECG signals.Among them,the study of the median filter shows that when the filter window is taken as 80%~120%of ECG signal cycle,the better filtering effect is obtained.At the same time,in this paper,based on the correlation of multiple leads,the separation of the effective signal and noise is carried out by using independent component analysis.That is,by calculating the correlation between the leads,the ECG signal containing only QRS waves are isolated and detected for the next step,thus reducing the interference of non-correlated lead noise.Then,the linear adaptive network instead of the original nonlinear neural network.This paper designed a nonlinear adaptive whitening filter,to improve the suppression of nonlinear noise in ECG signal,and use the error output of the network to enhance the R wave.In order to find the position of R wave accurately,this paper introduces the design of matched filter.Using the detected QRS wave as the template,we pass it through the whitening filter with the same parameters as the original ECG signal,then matched with the original signal filter.Through the experimental comparison,when select three layer neural network with 6 input nodes and 4 hidden nodes,we can get better performance to restrain nonlinear noise and the best detection accuracy.Finally,we design a dynamic threshold adjustment strategy based on the characteristic wave peak and RR interval probability distribution.It reduces the false detection and missed detection caused by traditional fixed threshold strategy.In the MIT-BIH arrhythmia database and St.-Petersburg database test,the positive predictive rate of this detection algorithm are 99.89%and 99.88%,and the false detection rate are 0.35%and 0.19%.Compared to other multi-lead detection algorithm,this algorithm has certain advantages.