| 摘要 |
PROBLEM TO BE SOLVED: To make compatible both reduction in musical noise and effective suppression of noise components. SOLUTION: A first noise-suppressing section 32 subtracts the spectrum Nw[j] of a stationary noise from the spectrum X[j] of an acoustic signal V[j] of each channel, coresponding to a subtraction coefficientα. A coefficient-setting section 44 generates a filter coefficient W, which emphasizes a target sound component from among spectra Nd[1]-Nd[J] of non-stationary noises. A second noise-suppressing section 42 executes directional array processing, to which the filter coefficient W is applied, to spectra Y[1]-Y[J], posterior to processing by the first noise-suppressing section 32 to generate a spectrum Z. An index-computing section 62 computes a kurtosis variational index KR, which shows the degree of a variation in kurtosis in the frequency distribution of signal intensity between, prior to the processing by the first noise-suppressing section 32 and posteriority to the processing by the second noise-suppressing section. A coefficient-adjusting section 64 variably controls the subtraction coefficientαso that the kurtosis variational index KR approaches- a target value K0. COPYRIGHT: (C)2011,JPO&INPIT
|
