Study of the Automatic Detection of Parkison’s Disease Based on Speaker Recognition Technologies and Allophonic Distillation

Abstract

The use of new tools to detect Parkinson´s Disease (PD) from speech articulatory movements can have a considerable impact in the diagnosis of patients. In this study, a novel approach involving speaker recognition techniques with allophonic distillation is proposed and tested separately in four parkinsonian speech databases (205 patients and 186 controls in total). The results of applying this new scheme in the databases provides up to 94% of accuracy in the automatic detection of PD and improvements up to 9% respect to baseline techniques. Results not only point towards the importance of the segmentation of the speech for the differentiation of parkinsonian and control speakers but confirm previous findings about the relevance of plosives and fricatives in the detection of parkinsonian dysarthria.

Study of the Performance of Automatic Speech Recognition Systems in Speakers with Parkinson’s Disease

“…Additionally, the analysis relative to ASR 1 suggests that the substitution rate of accented vowels (specially in the case of <í>) tends to be higher in patients, which is related to disprosody, commonly present in speakers with PD. Other characters that have a clearly higher substitution rate in patients compared to controls are <g, r, p, v> and <y>, findings which are in line with previous studies [34,32,35,6]. Fig.…”

“…Some studys suggest that 90% of PD patients suffer from dysarthria after 7 years since diagnosis [4]. However, although the influence of PD on the speech produced by patients with *Equal contribution PD is not always perceivable by human listeners, research using machine learning approaches has found PD-related cues in the speech of most of the studied patients, even those in the early developmental stages [5,6,7]. We therefore hypothesize that given that parkinsonian speech includes some specific traits even when no dysarthria or dysphonia is perceived by human listeners, these PD-related cues may influence the performance of Automatic Speech Recognition (ASR) systems.…”