Classification of heart signals: A nonlinear approach based on music effects

In this work the classification of heart signals based on nonlinear features is investigated. Physical responses to music, include variations in heart rate(HR). Indeed, the attempt to provide incontrovertible evidence of music induction on heart signals, remains a tremendous challenge due to thenoisy nonstationary nature of heart signals. To decompose and analyze thecyclic components of HR signals, empirical mode decomposition (EMD)is used as an adaptive mathematical tool, which extracts the recognizable and functional features for classification. Intrinsic mode function (IMF) values are calculated to determine whether the changes in signal features are experimentally significant due to the music. A set of valid features areproposed to serve as classifier input, circumventing the shortcomings of filtering methods. To determine the most efficient map between music andheart signals, the classification process is discussed with particular emphasis on the performance of three different classifiers: neural networks (NNs), Adaptive neuro fuzzy inference system (ANFIS), and Elman recurrent neural network (ERNN). Experimental performance over 62cases is reported. As the results indicate, the proposed method produces satisfactory classification accuracy and validates the generalizationcapability of proposed method. Generally NNs performed better than ANFIS. In classifying the maximum frequency (MaxFreq) and sampleentropy (SampEn) features, the best results are achieved by ERNN. Considering the maximum amplitude of fast Fourier transform (MaxFFT),MaxFreq and SampEn features as an input of the classifiers, feed-forward neural network (FFNN) has the best performance with the least errors, which proves the enormous efficiency of the NNs due to the application of Levenberg–Marquardt (LM) backpropagation algorithm

کلیدواژه ها:

Adaptive neuro fuzzy inference system (ANFIS) ، empirical mode decomposition (EMD) ، heart rate (HR) ، Iranian music ، classification