Factors affecting the intelligibility of synthesized speech

Abstract

Intelligibility measures for three speech synthesizers, the Echo IIb™, SmoothTalker™ (version 2.0, male voice), and DECTalk™ (“Perfect Paul”) were obtained using two methods of presentation of stimuli (audiotaped vs. live) and two methods of response (written vs. spoken). Twenty adult listeners each listened to 30 randomized synthesized sentences. Consistent with previous studies, the Echo IIb™ synthesized output was the poorest in terms of intelligibility; the DECTalk™ output was the best. The method of response did not appear to have an effect on the intelligibility measures. However, there was a significant difference between the methods of presentation. All three types of synthesized speech output were more easily understood in the live condition. Implications for future research are discussed.

• augmentative and alternative communication (AAC)
• intelligibility
• speech synthesis
• technology
• voice output communication aid (VOCA)