| 摘要 |
<p>Upon detecting an utterance period by a state decision part 14, a sound source position detecting part 15 detects the positions of sound sources 91 to 9K are detected by a sound source position detecting part 15, then covariance matrix of acquired signals are calculated by a covariance matrix calculating part 18 in correspondence to the respective sound sources, and stored in a covariance matrix storage part 18 in correspondence to the respective sound sources. The acquired sound level for each sound source is estimated by an acquired sound level estimating part 19 from the stored covariance matrix, and filter coefficients are determined by a filter coefficient calculating part 21 from the estimated acquired sound levels and the covariance matrices, and the filter coefficients are set in filters 121 to 12M. Acquired signals from the respective microphones are filtered by the filters, then the filtered outputs are added together by an adder 13, and the added output is provided as a send signal; by this, it is possible to generate send signals of desired levels irrespective of the positions of sound sources.</p> |
