Automatic Detection and Classification of Voice Pathology

Abstract

The result of rough vocal use is commonly voice pathology. Poor vocal practice can result in worse exceptional of voice, vocal fatigue, and vocal stress. This research utilizes glottal signal (signal produced by vocal folds) parameters to help out in identify voice disorders linked to vocal folds pathologies. For each recorded speech, the respective glottal signal is acquired. We select the most relevant as far as pathological / normal discrimination is concerned from the enormous set of parameters obtained. In this paper a new glottal signal parameter Maximum Opening Quotient (MOQ) is calculated to find Pathological / Normal speech discrimination. Using distinct options, the outcomes are compared. Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) algorithms are used for classifications. Result shows that the average efficiency rise 2.1% using the newly studied glottal parameter Maximum Opening Quotient (MOQ), which is a major contribution of this research.