摘要 |
PURPOSE:To provide a voice recognition device which emits the recognition result of high calculating accuracy at a high speed. CONSTITUTION:A voice signal division sensing part 3 senses the voice signal division on the basis of a characteristical quantity of information from an acoustic input signal, and an analytical processing part 4 calculates the mean of a self-correlation factor due to a short time analysis of the voice signal division a time series LPC kepstrum factor determined from it, and an LPC kepstorum factor in the voice signal division. A kepstorum factor subtracting part 5 subtracts the mean 4 of LPC kepstrum factors from the time series LPC kepstorum factor 4 of voice signal, and a standard voice pattern accumulation part 7 accumulates the result from the processing made by the analytical part 4 and subtracting part 5 about the voice having known contents of voice emission, and the voice pattern prepared about the fed voice of unknown contents of voice emission is put in reference to the standard voice pattern. |