Development of an Anthropomorphic Talking Robot - Development of Auditory Feedback Mechanism of Consonant Sounds

摘要

We have developed auditory feedback mechanism of an anthropomorphic talking robot WT-5(Waseda Talker No. 5) for clarifying human speech acquisition mechanism. In human speech acquisition, humans build phoneme category through mimic speaking. In our previous research, we have developed mimicking of consonant speech using sensory feedback and speech recognition. In this paper, we describe a consonant mimicking mechanism from auditory feedback. In this feedback mechanism, we use MFCC (Mel-Frequency Cepstrul Coefficients) and consonant time interval as acoustic features of consonant sounds. The robot parameters which control glottal and articulatory positions are determined so as to minimize the acoustic error between the human and robot speech.