Voice Disorder Detection Based on Automatic Speaker Identification Techniques

In this paper, we investigate the proprieties of automatic speaker identification (ASI) to develop a system for voice pathologies detection, where the models do not correspond to different speakers but it corresponds to classes of patients who share the same diagnostic. One essential part in this topic is the database (described later), the samples voices (healthy and pathological) are chosen from a German database which contains many diseases, spasmodic dysphonia is proposed for this study. A supervised algorithm is used to accomplish this task, Mel frequency cepstral coefficients (MFCCs) with first and second derivations are proposed as features, and modeled by weighted Gaussian mixture model (GMM) as it is used in ASI. The work is simulated using MATLAB, thanks to the toolbox voicebox for features extraction and dcpr for training and testing steps. The results are encouraging for further investigation on better classifiers. KeywordsVoice disorders, Speech patholodies detection, classifiction techniques, machine learning, laryngeal deseases.