Identification of normal and pain infants based on individual crying pattern

Abstract

An Infant informs his or her needs to those around them by crying. It is difficult for us adults to exactly know the message associated with each crying pattern. In this endeavour, a normal cry and a cry associated with pain will be identified using a signal processing approach. There are four processes involved; first stage is to filter the signal using pre-emphasis filter, then to perform feature extraction using Melfrequency cepstral coefficient (MFCC) and finally to classify the features into normal cry pattern and pain cry pattern using Radial Basis Function Neural Network (RBF). The accuracy achieved is 92.3%. Thus, the RBF has the potential to be utilized as a classifier for crying pattern signals.