Classification of Voice Pathology Using SVM Classifier

Abstract

Current research proposals have shown that voice pathology detection systems can significantly improve the assessment of voice disorders and provide early warning of voice pathologies. These systems made use of machine learning strategies, which are thought to be particularly effective instruments for identifying speech disorders. Nevertheless, the majority of suggested algorithms for detecting voice disorders used a small database. Low accuracy rate continues to be one of the most difficult problems for these procedures. This study describes a method for detecting vocal disorder using Support Vector Machine (SVM) to classify the voice signal into healthy or pathological samples. The vocal features in this study are Peaks, Zero Crossing Rate (ZCR), Fundamental Frequency (F0), Jitter, Max. Vowel /a/ speech samples were equally obtained from the Saarbrücken voice database (SVD). The classifier accuracy is around 92% after using these features.