On the development of an ASR-based multimedia game for speech therapy: Preliminary results

Abstract

A potential consequence of the ageing population is an increased incidence of neurological diseases that cause communicative disorders. In turn, this may lead to an increasing demand of intensive and costly speech therapy. To alleviate this problem, multimedia applications in the area of telerehabilitation and web-based speech training have been developed to support speech therapy. However, due to the repetitive nature of some exercises, therapy is not always perceived as particularly motivating. This paper reports on research aimed at developing a multimedia game that incorporates Automatic Speech Recognition (ASR) technology to provide patients autonomous and motivating practice without the intervention of a speech therapist. Currently, the game includes visual feedback on two dimensions of dysarthric speech that often deviate from healthy speech. To explore the possibility of integrating feedback on dysarthric speech by using ASR technology, initial experiments were conducted on available speech databases. The results show that employing ASR is becoming feasible thanks to recent developments in acoustic modelling.