Three-dimensional particle swam optimisation of Mel Frequency Cepstrum Coefficient computation and Multilayer Perceptron neural network for classifying asphyxiated infant cry

The performance Mel Frequency Cepstrum Coefficient (MFCC) in extracting significant feature is influence by several important parameter settings, namely the number of filter banks, and the number of coefficients used in the final representation. These settings affect the way the features are represented, and in turn, effect the performance of the classifier for diagnosis of the disease. Particle Swarm Optimization (PSO) algorithm is used in this work to adjust the parameters of the MFCC feature extraction method, together with the Multi-Layer Perceptron (MLP) classifier structure for diagnosis of infants with asphyxia. The extracted MFCC features were then used to train several MLP classifiers over different initialization values. The simultaneous optimization of MFCC parameters and MLP structure using PSO yielded 93.9% of classification accuracy.