Automatic Infant Cry Classification Using Radial Basis Function Network

Abstract

This paper proposes the automatic infant cry classification to analyse infant cry signals. The cry classification system consists of three stages: (1) feature extraction, (2) feature selection, and (3) pattern classification. We extract features such as Mel Frequency Cepstral Coefficients (MFCC), Linear Prediction Cepstral Coefficients (LPCC), and dynamic features to represent the acoustic characteristics of the cry signals. Due to the high dimensionality of data resulting from the feature extraction stage, we perform feature selection in order to reduce the data dimensionality by selecting only the relevant features. In this stage, five different feature selection techniques are experimented. In pattern classification stage, two Artificial Neural Network (ANN) architectures: Multilayer Perceptron (MLP) and Radial Basis Function Network (RBFN) are used for classifying the cry signals into binary classes. Experimental results show that the best classification accuracy of 99.42% is obtained with RBFN.