J Wave Signal Detection And Extraction Based On Wavelet Packet Transformation

Nowadays, with the rapid development of science and technology, therhythm of people’s work and life is becoming more and more fast, followed by apeople’s pressure increasing. In recent years, the incidence of cardiovasculardisease and its associated complications is higher and higher. J wave syndromeis one of these cardiovascular disease. Many studies have shown that, J wavesyndrome may cause a series of cardiac events (such as ventricular tachycardia,ventricular fibrillation and even sudden death). J point is a sudden change pointat the end of the QRS wave and the beginning of ST segment beginning. J waveis dome or hump shaped potential changes between QRS wave and ST segmentof electrocardiogram(ECG). The most important clinical significance of J wavesyndrome is to predict sudden cardiac death. Therefore, the correct detection of Jwave signal is crucial.To date, most research on J wave only at the clinical study stage, patient’scondition is diagnosed by doctor’s experience on ECG. These diagnosis islimited to the signal amplitude, waveform and position judgment, and it does notform a research system. Especially when J wave amplitude is low, not easily observed by naked eyes, it often cause misdiagnosis. Study ECG at theperspective of signal processing, it can amplify the waveform in time domainsignal processing methods, and study characteristics both in time and frequencydomain. It can improve the ability of detecting J wave and identify high-riskpatients with abnormal J wave in clinical.first of all, as study of J wave in signal processing field is blank, weresearch its classification standard. This paper introduce the basic knowledge ofwavelet (packet) analysis, decompose three kinds of different signal (normalECG signal, N-J signal and J wave signal) based on wavelet packet. Thenreconstruct3D figure. X axis,y axis and z axis represents time,scale and energydistribution respectively. Extract the feature of three different3D figures,wemade the make the classification criteria finally.Secondly, we separate signals mixed with J wave signal, normal ECGsignal and noise signal based on sparse component analysis. Due to theinsufficient of sparse, traditional sparse component analysis algorithm can’twork well. In this paper, we improve the classic algorithm and separate threesource signals step by step successfully. Compared with fast independentcomponent analysis algorithm shows that proposed algorithm in this paper hashigher accuracy.