A Comparative Study on Dysophonia Classification

The human voice production system is a complex biological apparatus that can change pitch and volume. Internal and external forces frequently damage the vocal folds, resulting in a change in voice. The automated detection of voice abnormalities using machine learning algorithms is important since it has been shown to make the process of perception of the disorder easier. In which, the features of the voice signal plays a key role to achieve valid accuracy. This paper offers a study of PLCC, DWT, MFCC and GFCC feature set for classification of pathological voice signals. SVM, KNN and Decision tree are the classification technique used for analysis. The results obtained for the dataset used implementation shows that a combination of GFCC features with SVM classification obtained an accuracy result of 98.7%.