How Dysarthric Prosody Impacts Naïve Listeners’ Recognition

Autor: Verkhodanova, Vass, Timmermans, Sanne, Coler, Matt, Jonkers, Roel, de Jong, Bauke, Lowie, Wander, Salah, Albert Ali, Karpov, Alexey, Potapova, Rodmonga
Přispěvatelé: Culture, Language & Technology, Cancer Research Center Groningen (CRCG), Movement Disorder (MD)
Jazyk: angličtina
Rok vydání: 2019
Předmět:
Zdroj: Speech and Computer: 21st International Conference, SPECOM 2019, Proceedings, 510-519
STARTPAGE=510;ENDPAGE=519;TITLE=Speech and Computer
Speech and Computer ISBN: 9783030260606
SPECOM
Popis: The class of speech disorders known as dysarthria arise from disturbances in muscular control over the speech mechanism caused by damage of the central or peripheral nervous system. Dysarthria is typically classified into one of six classes, each corresponding to a different neurological disorder with distinct prosodic cues [3]. The assumption in this classification is that dysarthric speech can be classified implicit on the basis of perception. In this study, we investigate how accurately naïve listeners can recognize stress and intonation in dysarthric speech, and if different neurological disorders impact the ability to convey meaning with these same two cues. To those ends, we collected speech data from Dutch speakers diagnosed with cerebellar lesions (ataxic dysarthria), Parkinson’s Disease (hypokinetic dysarthria), Multiple Sclerosis (mixed classes of dysarthria) and from a healthy control group. Thirteen naïve Dutch listeners participated in the perceptual experiment which targeted recognition of intended realization of four prosodic functions: lexical stress, sentence type, boundary marking and focus. We analyzed recognition accuracy for different groups and performed acoustic analyses to check for fundamental frequency trajectories. Results attest to different accuracy recognition results for different disease groups. The sentence type recognition task was the most sensitive of all tasks for differentiating different diseases both on perceptual and acoustic levels of analysis.
Databáze: OpenAIRE