Glottal signal parameters as features set for neurological voice disorders diagnosis using K-Nearest Neighbors (KNN)

Disorders affecting nervous system can affect the voice in different ways. Different neurological disorders may lead to speech problems; this may modify the articulatory characteristics related to vocal folds function, which provide important information for detecting certain neurological diseases. In order to improve the diagnosis of Neurological Voice Disorders (NVD) an objective technique based on articulatory evaluations of vocal folds vibration, based on an estimation of a Glottic Signal (GS) extracted from Speech signal. In this work, we propose a method based on parameters extracted from GS obtained by an inverse filtering algorithm for automatic classification and diagnosis of NVD using K-Nearest Neighbors (KNN). Our work is developed around Saarbrucken Voice Database it is an open German database containing deferent samples, words, sentences of normal and pathological voices. We have selected three groups of subjects: persons with normal voices, which considered as reference, persons having suffered Parkinson disease (PD) and persons with spasmodic dysphonia.