Glottal Signal Analysis for Voice Pathology

The work present in this paper is focusing on a method for the detection of voice disorders related to the pathologies of vocal folds. It extracts glottal signal parameters by using inverse filtering method. Support Vector Machine (SVM) and k-nearest neighboring (k-NN) classifiers are used for differentiating healthy and non-healthy persons. The voice samples of the person belonging to both the categories were acquired using a specialized voice recorder. SVM provided 98.5 % whereas k-NN provided 88.2 % accuracy. Voice signals of patients affected with vocal cord nodules were also collected and tested with same classifier. The Accuracy of 78.9 % and 70.4% were obtained with SVM and k-NN respectively.