No abstract

No abstract

No abstract

No abstract

IntroductionFrequency-amplitude parameters of ECG signals are rather sophisticated. Analysis of ECG signals requires high frequency and amplitude resolution. This is required to localize low frequency and high frequency compo nents.There are two methodological approaches to analysis of this problem.The first approach is preliminary division of nonsta tionary signal into time segments (frames). The time seg ments are quasi stationary portions with invariable statis tics (almost invariable) within a given time range and fur ther parametric analysis (autoregression model AR [1] or sliding mean model [2] or local Fourier transform). Quasi stationary portions should be detected with mini mal delay. Series statistical analysis methods are used to detect fast discoordination moments [3,4].The second approach includes wavelet transform. The signal is decomposed using basis functions. Methods of wavelet analysis allow frequency-time heterogeneity of signal to be detected without ergodicity limitation [5]. In contrast to Fourier transform (FT), wavelet transform (WT) provides 2 D presentation of signal within a fre quency area. The argument is basis function (time) and position is characterized by basis function shift. Large sig nal elements can be separated from small signal elements by localizing on the time scale. WT is a local spectral analysis [6].ECG analysis using continuous WT (CWT) was dis cussed in [7 9]. Numerical CWT is rather difficult and involves [10, 11] time scale selection, realization length selection, border effect in wavelet spectrum, WT scale ratio to frequency, and CWT accuracy in numerical analysis. CWT accuracy in numerical analysis was dis cussed in [12]. In the general case, the CWT accuracy in numerical analysis for functions with non continuous derivatives is insufficiently understood.In this work spectral analysis of electrocardiographic signals based on wavelet packet processing was used [13,14].This approach is superior to FT and WT spectral analysis [15]. ECG signal was identified using mean power of wavelet packet coefficients.CWT can be used in noninvasive electrocardiogra phy for diagnosis of such cardiovascular system disorders as ciliary arrhythmia, ventricular tachycardia (VT), and ventricular fibrillation (VF). WTA provides another numerical parameter (nodal entropy of the ECG tree). This parameter also has diagnostic value [13]. Spectral ECG Analysis Using Wavelet BasisIn terms of digital filtration, WT is implemented as two filters with finite pulse characteristic (FPC) with fur ther time decimation: signal is filtered using low fre quency (LF) and high frequency (HF) filters at cut fre quency π. This decreases the frequency range and sam pling frequency of the HF and LF components by one half. Both the HF and LF components are stored in computer memory. This scheme is called subcavity cod ing scheme [16].FPC filters give rise to aliasing effect. Rectangular filters avoid aliasing effect but induce Gibbs effect and signal discontinuity at terminal fragments. Weight func ECG signals...