| 摘要 |
PROBLEM TO BE SOLVED: To provide a speech recognition device enabling a higher speech recognition rate by estimating a direction or a position of a speaker. SOLUTION: In the speech recognition device provided with a microphone array 10 comprising plural microphones juxtaposed at prescribed intervals, a direction estimation part 13 estimates an azimuth of at least one of sound sources received by the microphone array based on the electric signals outputted from each microphone, and a beam forming part 14 generates at least one beam signal corresponding to the direction of the azimuth of at least one sound source estimated on the basis of the electric signals outputted from each microphone. Next, based on each beam signal, a sound source judging part 16 judges whether each beam signal is voiced speech or unvoiced speech, using HMM of speech and HMM of noise, and when the beam signal is judged as speech, a speech recognition part 17 performs speech recognition operation to the beam signal and outputs a speech recognition result. |
