L. Abou-Abbas,
C. Tadj, C. Gargour, L. Montazeri
This
paper addresses the problem of automatic cry signal segmentation for
the purposes of infant cry analysis. The main goal is to automatically
detect expiratory and inspiratory phases from recorded cry signals. The
approach used in this paper is made up of three stages: signal
decomposition, features extraction and classification. Different
decomposition techniques such as Short Time Fourier Transform,
Empirical Mode Decomposition and Wavelet Packet Transform have been
considered and various set of features have been obtained. Supervised
classification approaches such as GMM and HMM with 4 and 5 states have
been discussed as well. The experiments have been repeated several
times for different training and testing data sets, randomly chosen
using a 10-fold cross validation procedure. Global classification error
rates of around 8.9% have been achieved.
Keywords: Automatic
Segmentation, Empirical Mode Decomposition, Wavelet Packet Transform,
Mel-Frequency Cepstral Coefficients, Gaussian Mixture Models, Hidden
Markov Models..