| 摘要 |
The method involves using a bypass factor (alpha) with a value between 0 and 1 to reduce the effect of errors in the estimated value of the power density spectra of a speech signal Ps(f) and a noise signal Pn(f). The bypass factor constant, alpha, is replaced by an adaptively controlled bypass factor alpha(w), and an additional attenuation factor beta(w) is used, where w corresponds to the probability of a speech signal being present. The transfer function is determined from HAB(f)=(Ps(f)+alpha(w)Pn(f))/(Ps(f)+Pn(f)) .beta(w) An Independent claim is included for a noise filtering apparatus. |
