Evaluation of an Automatic Speech Recognition Platform for Dysarthric Speech

• Published in: Folia phoniatrica et logopaedica Vol. 73; no. 5; pp. 432 - 441
• Main Authors: Calvo, Irene, Tropea, Peppino, Viganò, Mauro, Scialla, Maria, Cavalcante, Agnieszka B., Grajzer, Monika, Gilardone, Marco, Corbo, Massimo
• Format: Journal Article
• Language: English
• Published: Basel, Switzerland Karger 01.09.2021
• Subjects: Cognitive Sciences; Computation and Language; Computer Science; Dysarthria - diagnosis; Humans; Life Sciences; Neurons and Cognition; Research Article; Speech; Speech Intelligibility; Speech Perception; Speech Production Measurement; Speech Recognition Software; Disabilities and handicaps; Dysarthria; Intelligibility; Automatic speech recognition; Speech therapy; Automatic Speech Recognition; mPASS platform; Automatic Speech Recognition dysarthric speech mPASS platform neurological disorder; dysarthric speech; neurological disorder
• ISSN: 1021-7762; 1421-9972
• DOI: 10.1159/000511042

Introduction: The use of commercially available automatic speech recognition (ASR) software is challenged when dysarthria accompanies a physical disability. To overcome this issue, a mobile and personal speech assistant (mPASS) platform was developed, using a speaker-dependent ASR software. Objective: The aim of this study was to evaluate the performance of the proposed platform and to compare mPASS recognition accuracy to a commercial speaker-independent ASR software. In addition, secondary aims were to investigate the relationship between severity of dysarthria and accuracy and to explore people with dysarthria perceptions on the proposed platform. Methods: Fifteen individuals with dysarthric speech and 20 individuals with nondysarthric speech recorded 24 words and 5 sentences in a clinical environment. Differences in recognition accuracy between the two systems were evaluated. In addition, mPASS usability was assessed with a technology acceptance model (TAM) questionnaire. Results: In both groups, mean accuracy rates were significantly higher with mPASS compared to the commercial ASR for words and for sentences. mPASS reached good levels of usefulness and ease of use according to the TAM questionnaire. Conclusions: Practical applicability of this technology is realistic: the mPASS platform is accurate, and it could be easily used by individuals with dysarthria.