The methodology of ECG feature extraction based on empirical mode decomposition

Zhiqiang Zhao, Jianjun He, Jia

Abstract

It’s essential for authentication to extract the feature of individual ECG signal accurately and efficiently. The common time-frequency analyses such as the classical Fourier transform, wavelet transforms and adaptive decomposition deem Fourier Transform as the ultimate theory. However, due to the paradox when analyzing the non-stationary signal, like false signals production and aliasing, so it desires for improvement for ECG signal feature analysis. The EMD (Empirical Mode Decomposition) is completely irrelevant to the Fourier analysis framework and it’s effective to extract non-stationary and nonlinear signals. The ECG signal is decomposed into a series of basic mode components and a remainder term, and the acquisition of each basic mode of IMF depends entirely on the local time scale of signal without using any information. Improved EMD algorithm and the improved mirror extension method inhibited the use of end effect, improving the accuracy of the EMD.