Hypothyroidism in infants is caused by the insufficient production of hormones by the thyroid gland. Due to stress in the chest cavity as a result of the enlarged liver, their cry signals are unique and can be distinguished from the healthy infant cries. This study investigates the effect of feature selection with Binary Particle Swarm Optimization on the performance of MultiLayer Perceptron classifier in discriminating between the healthy infants and infants with hypothyroidism from their cry signals. The feature extraction process was performed on the Mel Frequency Cepstral coefficients. Performance of the MLP classifier was examined by varying the number of coefficients. It was found that the BPSO enhances the classification accuracy while reducing the computation load of the MLP classifier. The highest classification accuracy of 99.65% was achieved for the MLP classifier, with 36 filter banks, 5 hidden nodes and 11 BPS optimised MFC coefficients.
This paper presents a new application of the Particle Swarm Optimization (PSO) algorithm to optimize Mel Frequency Cepstrum Coefficients (MFCC) parameters, in order to extract an optimal feature set for diagnosis of hypothyroidism in infants using Multi-Layer Perceptrons (MLP) neural network. MFCC features is influenced by the number of filter banks (f(b)) and the number of coefficients (n(c)) used. These parameters are critical in representation of the features as they affect the resolution and dimensionality of the features. In this paper, the PSO algorithm was used to optimize the values of f(b) and n(c). The MFCC features based on the PSO optimization were extracted from healthy and unhealthy infant cry signals and used to train MLP in the classification of hypothyroid infant cries. The results indicate that the PSO algorithm could determine the optimum combination of f(b) and n(c) that produce the best classification accuracy of the MLP.
Combined Support Vector Machine (SVM) and Principal Component Analysis (PCA) was used to recognize the infant cries with asphyxia. SVM classifier based on features selected by the PCA was trained to differentiate between pathological and healthy cries. The PCA was applied to reduce dimensionality of the vectors that serve as inputs to the SVM. The performance of the SVM utilizing linear and RBF kernel was examined. Experimental results showed that SVM with RBF kernel yields good performance. The classification accuracy in classifying infant cry with asphyxia using the SVM-PCA is 95.86%.