Intelligent classification of ECG signals to distinguish between pre and on-music states

In this work, the classification of heart signals affected by music is investigated. The nonlinear and chaotic nature of ECG signals makes it desirable to develop and apply an intelligent mechanism for efficient signal classification. Afterwards, extracting the recognizable and functional features plays a significant role in classification accuracy. Empirical mode decomposition (EMD), as an adaptive mathematical analysis is applied to decompose the signals into a sum of components each called an intrinsic mode function (IMF). IMF values are applied to determine whether the changes in signal features are experimentally significant due to the music. The performance of two practical classification methods is reported to determine the most efficient input-output relationship between music and heart signals. Experimental results over 62 cases, validates the generalization capability of the proposed method and perform acceptable values of MSE for the classification process. Elman recurrent neural network (ERNN) performed most effectively in classifying the maximum frequency (MaxFreq) and sample entropy (SampEn) of IMF (2). However, results reflect the considerable potential of feed-forward neural network (FFNN) for the classification of maximum amplitude of FFT (MaxFFT) and MaxFreq of IMF (1).