An accurate infant cry classification system based on continuos Hidden Markov Model

This paper describes the feasibility study of applying a novel continuous Hidden Markov Model algorithm as a classifier to an automatic infant cry classification system which main task is to classify and differentiate between pain and non-pain cries belonging to infants. The classification system is trained based on Baum - Welch algorithm on a pair of local feature vectors. In this study, Mel Frequency Cepstral Coefficient (MFCC) and Linear Prediction Cepstral Coefficients (LPCC) are extracted from the audio samples of infant´s cries and are fed into the classification module. The system accuracy reported in this study varies from 71.8% up to 92.3% under different parameter settings, whereby in general the system that are bases on MFCC features performs better than the one that utilizes LPCC features. The encouraging results demonstrate that indeed Hidden Markov Model provides for a robust and accurate infant cry classification system.