Classification of Psychogenic and Laryngeal Voice Diseases Based on Teager Energy Operator

Among several ways of communication, the voice remains the fastest natural tool for human-to-human and human-to-machine communication. That is why the research in automatic voice pathology detection and classification area has gained much interest in the recent years. Indeed, these automatic systems may be considered as assistive tools for the physicians during the assessment stage. This latter may help them to make decision, whether the voice signal belongs to a healthy or unhealthy subject and identifies the nature of pathology. In this context, this paper provides a voice pathology detection and classification system based on wavelet analysis and Teager Energy Operator (TEO). First, we used the input voice signal that we taken from Saarbrucken Voice Database (SVD) [1], to extract a set of features. These feature vectors are fed into a Gaussian Mixture Model (GMM) [2] for the sake of classification. The obtained results are 96.66% for the detection task and 92.5 % using TEO. These results show that our proposal outperforms some state-of-art methods used in voice pathology identification.