Spectrogram analysis of ECG signal and classification efficiency using MFCC feature extraction technique

Abstract

This article focuses on ECG signal recognition based on acoustic feature extraction techniques. The SVM and k-NN classification approaches are proposed for recognizing the ECG heart sound as well as for calculating the recognition efficiency. In this proposed technique, ECG signals are previously transformed into a successive series of Mel-frequency cepstral coefficients for computing the acoustic features in terms of mean value. A histogram based understandable and new approach is proposed at this point for recognition of ‘P’ wave, ‘R’ wave etc. from ECG waveform. The recognition of ECG signal and their distinguishing features provide significant effort for the analysis. Here three statistical data with their detection efficiency estimation of histograms is analyzed from ECG signals from database. The entire method has been applied for convenience to different ECG record files taken from MIT-BIH database. Twelve leads are used from multi-lead ECG database which contains a 3600 Hz sampling frequency. The entire algorithm is executed on MATLAB R2014a. In this, the proposed method performance efficiency is evaluated.