A Comparative Study Of Parallel Multi-Layer Perceptron Machine Learning Models For Voice Pathology Detection

Abstract

The main objective of this paper is to develop a mobile healthcare framework that evaluates two multi-layer perceptron models for the detection of voice disorders. In particular, identifying healthy and unhealthy voices is a challenging task for physicians. Hence the design of an Automatic Mobile Healthcare System is investigated to detect normal and pathological voices using two different Convolutional Neural Network (CNN) models such as Alex-Net and Res-Net (Residual Neural Network). The voice samples were taken from the SVD database and finally, we correlate them, to draw a conclusion on which models perform well. As per the investigations carried out in this research work the Res-Net model with an accuracy of 85.29% outperforms the Alex-Net model which has an accuracy of 66.91% for detection of pathological voices.