The Research Of J Wave Extraction Technology Based On Sparse Blind Source Separation

Coronary artery disease is threaten to human health and it will be mostimportant that predicting disease for high-risk patient timely and exactly.Ventricular tachycardia may lead to lose people’s lives and ECG of patients beformed J wave. J wave formed by J point offset has low amplitude and if itmixes with normal ECG, it will not be observed easily. How to extract J waveshidden in ECG from observed signals and research and classify them in thefuture to provide basis for diagnosing Coronary heart disease in clinical hasgreat value in medical area. BSS is a good choice to solve the problem.At present, sparse sources are researched by scholar at home and abroadand many algorithms are proposed to separate sparse sources. Actually, becausesources are not fully sparse, the mixing matrix can not be estimated exactlybased on traditional SCA so as to extract J wave from the observed signalsincluding ECG and J waves. Estimating matrix and recovering sources madeBSS progressed and they are need to be solved quickly when the sources are notfully sparse.On the condition, this paper analyzed the basic theory of BSS andestimating mixing matrix and recovering sources based on traditional SCA indetail. The feedback partial sparse component analysis is proposed in this paperbased on the extracted perfect source by traditional SCA and K-means. Becausethe sources are overlapped, the extremum point of the overlapped areas is located and the pairs of points with same value are searched. Then the mixingmatrix can be estimated easily and exactly. Or after the sparse points are selectedand the ratio matrix is structured, the FPSCA is used to estimate the mixingmatrix. The experiment results show that the mixing matrix can be obtained andJ wave may be extracted exactly.