Toward the Automatic Generation of Cued Speech

摘要: Although Manual Cued Speech (MCS) can greatly facilitate both education and communication for the deaf, its use is limited to situations in which talker, or a transliterator, able produce cues cue receiver. The availability of automatically produced would substantially relax this restriction. However, it unclear whether current automatic speech recognition (ASR) technology be adequate producing automatically. To evaluate adequacy, we measured reception scores achieved by highly experienced receivers MCS when were one actual several hypothesized ASR systems, as well with speechreading alone MCS. systems studied modelled effects various types errors delays associated process, included representative state-of-the-art, speaker-dependent phonetic recognizer. Results indicate that while speaker-independent probably not provide useful cues, provided aid substantially. benefit generated heavily dependent on an effective visual display minimizes so are perceived synchrony facial actions.