摘要 |
<p>PROBLEM TO BE SOLVED: To provide a method and a device for speech recognition which can hold recognition precision in speech recognition excellent without lowering the output of an audio device so much. SOLUTION: A control part 15 selects an (N)th pattern among a plurality of equalizer patterns stored in an equalizer pattern storage part 22 in speech recognition processing. Consequently, data of the (N)th pattern are inputted to a graphic equalizer 21, which attenuates the level of a signal component in a 100 to 5 kHz frequency band used for the speech recognition larger than signal components in other frequency bands.</p> |