Telephone Communication: Synthetic and Dysarthric Speech Intelligibility and Listener Preferences

Abstract

For many individuals with severe speech intelligibility challenges, deciding whether to use one's own oral dysarthric speech or synthesized speech when communicating on the telephone is a difficult process. In the present study, we investigated the intelligibility of dysarthric speech and two types of synthesized speech over the telephone and in quiet conditions, and listeners' preferences for these speech types. Participants heard short sentences in one of three types of speech: dysarthric female speech, DECTalk® Beautiful Betty, and MacinTalk™ Victoria, high quality. They were asked to transcribe sentences for a measure of intelligibility, and to answer questions to indicate their preferences for speech type and their preferences for synthesizer. The overall intelligibility of DECTalk® and MacinTalk™ were significantly higher than the intelligibility of the dysarthric speech. Additionally, the intelligibility of the speaker with dysarthria decreased significantly when listeners were presented with the materials over the telephone. The two synthesized speech types (DECTalk® and MacinTalk™) did not differ significantly in intelligibility between the two conditions. On average, listeners tended to be more comfortable when listening to synthesized speech than when listening to dysarthric speech. Listener preferences for synthesizers were dependent on the condition: most listeners in the Speaker Group preferred the DECTalk® female voice to the MacinTalk™ female voice, while listeners in the Telephone Group were equally distributed.