摘要 |
<p>The echo-plus-speech input signal is canceled by transforming the input signal into spectral domain and decreasing the gain in each frequency which has a low signal-to-echo ratio. The gain in each frequency is controlled by the input signal-to-echo ratio estimate, modified by the prior input signal-to-echo estimate, a non-linear echo suppression function, and a smoothing function of the residual output echo signal estimate. <IMAGE></p> |