| 摘要 |
PURPOSE:To improve a recognition rate by multiplying the degrees of similarity and peak values obtd. from various sets of speech data obtd. from plural microphones to determine the degree of selection and outputting the final recognition result in accordance with the max. degree of selection. CONSTITUTION:To register speech data, a user sets a mode select switch 8 at registration and utters voices toward a microphone 1b. The speech signal is preserved via an amplifier 2b, a filter bank 3b, an A/D converter 4b, and a recognition section 5b into a data memory 6b and is transferred via a CPU 7 to memories 6a, 6c. To recognize the speech, the user selects a recognition mode with the switch 8. A CPU 7, then, sets speech recognition sections 5a, b, c, to wait for input. The respective recognition sections make DP matching with the registered data and multiply the degrees of similarity and the peak values from peak-hold circuits 9a, b, c. The recognition sections output the prescribed external control signal corresponding to the code number which emits the max. degree of selection. |
