Analysis of normal and pathological infant cries using bispectrum features derived using HOSVD

Abstract

In this paper, bispectrum-based feature extraction method is proposed for classification of normal vs. pathological infant cries. Bispectrum is a class of higher order spectral analysis, Bispectrum is computed for all segments of normal as well as pathological cries. Bispectrum is a two-dimensional (i.e., 2-D) feature. A tensor is formed using these bispectrum features and then for feature reduction, higher order singular value decomposition theorem (HOSVD) is applied. Our experimental results show 98.94 % average accuracy of classification with support vector machine (SVM) classifier whereas baseline features, viz., Mel frequency cepstral coefficients (MFCC), perceptual linear prediction coefficients (PLP) and linear prediction coefficients (LPC) gave classification accuracy of 53.99 %, 63.14 % and 63.07 %, respectively. High classification accuracy of bispectrum can be attributed to its ability to capture nonlinearity in the signal.