| 摘要 |
PROBLEM TO BE SOLVED: To improve noise reduction effect by improving computing precision. SOLUTION: Gains of amplifiers 13a to 14b, which amplify output signals of first and second microphones 11 and 12, are switched so that the gains are increased during a learning and decreased during a voice recognition operation. Thus, noise level during the learning and the speaker voice level during voice recognition are made optimum levels for AD converters 18 and 19. An adaptive signal processing section 20 conducts an adaptive signal processing by making the signals, which are outputted from the AD converter responding to one of the microphones during a learning, as target signals and the signals, which are outputted from the AD converter responding to the other microphone, as reference signals to determine the coefficients of an adaptive filter 20b. The filter coefficients are set in the adaptive filter during voice recognition and a subtracting section 22 outputs voice signals having a large SN ratio to a voice recognition section. |
