Text-to-speech (TTS) synthesis technology has great potential to be implemented in aural rehabilitation because of the flexibility it affords for stimulus generation. However, little is known about how listeners with cochlear implants (CI) perceive and respond to synthetic speech. This study investigated how adult CI users perceive and respond to synthetic speech of varying qualities in order to determine if they respond to synthetic speech in a different manner than individuals with normal hearing (NH). Sentence recognition performance of 16 adult CI users and 30 NH listeners was compared on a semantically unpredictable sentence task using synthetic speech generated at different levels of quality. The results indicated that while the CI group had an overall lower performance than the NH group, their pattern of responding was similar to NH listeners across the speech stimuli. These parallel sentence recognition results of synthetic stimuli in both groups are promising for the application of a concatenative TTS synthesis system for generating speech training material in auditory rehabilitation software for CI users.

Index Terms: speech synthesis, semantically unpredictable sentences, cochlear implants, speech perception, intelligibility, aural rehabilitation