ECG Diagnostic Using Multi-Scale Principal Components Analysis

In this paper, a method of ECG diagnostic is proposed. This method is based on three parts: the simplification using multi-scale PCA, the default detection and localization by classic linear PCA. The first part consists of applying a multi-scale analysis based on continuous wavelet transform (CWT); it allows to decompose the signal into wavelet coefficients on five levels of multi-scale resolution. Then, we apply the principal component analysis PCA wholes on the coefficients obtained to determine the number of principal components to retain and which provide more information. The signal is reconstructed by referring only scales of resolution whose main components are found. In the second part, we introduce the resulting signal for detecting and locating defects by PCA by introducing two statistics: SPE and T ². Variables are determined defective by the method of calculating contributions by the same previous statistics. Comparing the results obtained by this approach and the data of the expert proves its reliability in the diagnosis of ECG signal.