A Study on the Recognition of Patterns of Infant Cry for the Identification of Deafness in Just Born Babies with Neural Networks

摘要

In this paper we present the methodologies and experiments followed for the implementation of a system used for the automatic recognition and classification of patterns of infant cry. We show the different stages through which the system is trained to identify normal and hypo acoustic (deaf) cry. The cry patterns are represented by acoustic features obtained by the Mel-Frequency Cepstrum and Lineal Prediction Coding techniques. For the classification we used a feed-forward neural network. Results from the different methodologies and experiments are shown, as well as the best results obtained up to the moment, which are up to 96.9% of accuracy.