Pathological voice discrimination using cepstral analysis, vector quantization and Hidden Markov Models

Abstract

Pathological voice discrimination has been made using digital signal processing techniques as a complementary tool to videolaringoscopy exams. This method is non-invasive to patients compared to laringoscopy. This paper aims at analyzing the use of cepstral analysis to discriminate voices affected by vocal fold pathologies. A Vector Quantizer using a distortion measurement followed by a Hidden Markov Model-based classifier is employed. Results obtained show an effective and objective way in analyzing voice disorders caused by a vocal fold pathology.