A Survey on ASR Systems for Dysarthric Speech

Abstract

Recently Automatic Speech Recognition (ASR) has been widely overblown with many applications and assistance but orally challenged people, such as people with disordered speech, can’t get much benefits. Speech technologies are very useful on a daily basis to assist people with speech disorders. Dysarthria is a neurological speech disorder caused by significant injury in the left hemisphere of the brain. Dysarthric people have difficulty in the movement of speech-related muscles. As a result of strain on their speech muscles, individuals with dysarthria are able to generate limited speech data for analysis.In order to recognize speech of dysarthria sufferers, a robust technique is needed that can cope with extreme irregularity and narrow training data. This survey details a brief understanding of dysarthric speech characteristics and behavior. It also presents several attempts that have been made to make robust ASR systems for dysarthric speech.