Automatic Diagnosis of Asphyxia Infant Cry Signals Using Wavelet Based Mel Frequency Cepstrum Features

Authors:

Abstract
A powerful automatic recognition system can have a great impact in decreasing infant mortality by early diagnosing different pathologies affecting newborn life. In this paper, an automatic infant cry classification model is developed to distinguish between normal and asphyxiated infants. This research studies the performance of using discrete wavelet derived Mel frequency cepstrum coefficients in estimating features from infant cry signals. The extracted features are fed to a binary support vector machine classifier and classification accuracy is computed to examine the efficiency of developed model compared to conventional Mel frequency cepstrum feature estimation technique. Results show that using wavelet derived Mel frequency cepstrum extracted features has produced a higher classification accuracy of 98.5% in discriminating normal and asphyxia infant cry signals.

This publication has 11 references indexed in Scilit: