Wavelet neuron selection method for ECG data compression

Abstract

In this paper, a wavelet network for the Electrocardiograph (ECG) data compression and the selection of its wavelet neuron are presented. The methods of the frequency-domain matching and the orthogonal least square (OLS) algorithm in selecting the wavelet basis and its quantity were discussed. We choose Morlet wavelet as the mother wavelet, and use the ECG signal for simulation. The result demonstrates that the number of Morlet wavelets whose spectrums locating at the ECG is up to 152. But after filtrated by the OLS algorithm, it reduces sharply. This method can make the size of the wavelet network driving to optimum and reduce the training time of the wavelet network significantly. The algorithm also can reconstruct the ECG signal very well. The results of simulation indicate that it can reflect the location and intensity of all waves correctly. Consequently, the algorithm has higher compression ratio and fidelity.