Research on the application of empirical mode decomposition algorithm in human medical signal
Due to the nonlinear and non-stationary characteristics of biological signals,some traditional methods such as Fourier transform may be difficult to accurately capture the multi-scale structure of the signal,resulting in inadequate feature extraction.In this paper,a combined approach of ensemble empirical mode decomposition and multi-scale permutation entropy is proposed.Firstly,by applying the adaptive noise complete set empirical mode decomposition algorithm(CEEMDAN),the signal can be efficiently decomposed into multiple eigenmode functions(IMFs).Secondly,multi-scale permutation entropy is introduced to analyze the complexity of these IMFs in order to study the change of signal complexity at different time scales.Then,the dynamic change characteristics of signals at different time scales are input into SVM classifier,and the results are divided into normal and abnormal types.Finally,the combined method is applied to different types of human medical signals,including heart sound signal and EEG signal.The experimental results show that the CEEMDAN algorithm combined with multi-scale permutation entropy can effectively improve the classification accuracy.The average accuracy of EEG epilepsy signal classification was 99.8％,and the performance index of heart sound signal classification model was 94.71％.
万方数据
维普
