Understanding the Biological Underpinning of Auditory Perception for Vowel Sounds Using a Type-2 Fuzzy Neural Network

摘要

The temporal lobe in the human brain is responsible for low-level audio perception, whereas the pre-frontal lobe takes active role in interpreting the audio information. This paper introduces a novel approach to understand the interrelation between the temporal and the pre-frontal lobes of the brain in interpreting vowel sounds. The inter-relation is ascertained by two approaches. The first approach computes correlation measure between the direct brain signals of the said two lobes. The higher the correlation coefficient, the better is the interrelation between the activated lobes. The second approach aims at developing a feature-level mapping between the temporal and the prefrontal lobe brain activations. The motivation of the second approach lies in examining the uniformity in the learnt neural weights after convergence for the same vowel audio stimulus irrespective of the diurnal variations in the brain signals. Although any traditional mapping functions could be utilized to undertake the temporal to prefrontal mapping, we used a type-2 fuzzy neural network to serve the purpose. Experiments undertaken confirm that the weights of the proposed type-2 fuzzy neural net converges faster than its type-1 counterpart and back-propagation neural network. The faster convergence of weights represent that the proposed type-2 fuzzy neural network captures better audio perceptual ability than the rest. The proposed work is expected to find applications in the early detection of disorder in auditory perceptual-ability, usually referred to as Dyslexia.