Lipreadability of a synthetic talking face in normal hearing and hearing-impaired listeners

摘要: The Synface project is developing a synthetic talking face to aid the hearing-impaired in telephone conversation. This report investigates gain intelligibility from head when controlled by hand-annotated speech both 12 normal hearing (NH) and 13 (HI) listeners (average loss 86 dB). For NH listeners, audio everyday sentences was degraded simulate information losses that arise severe-to-profound impairment. HI group, filtered speech. Auditory signals were presented alone, with face, video of original talker. Purely auditory low for group. With addition average increased 22%. group had large variation purely condition. They showed 22% improvement face. groups, significantly lower than natural However, sufficient be useful communication. Questionnaire responses indicated strong interest system.