| 摘要 |
PURPOSE:To surely detect double talking and to improve an echo elimination rate. CONSTITUTION:A sound signal encoded by a CELP system is decoded by a decoder 13, the decoded signal is supplied to a pseudo echo line 24 and an impulse response estimating part 25, D/A converted and then supplied to an echo line 23. A signal from the echo line 23 side is A/D converted, a signal outputted from the pseudo echo line 24 is subtracted from the A/D converted signal by an erasing circuit 18 and the result is supplied to a sound encoder 19 and also to the estimating part 25. From a spectrum envelope decoder 47 in the decoder 13, a spectrum envelope and pitch period power are extracted, which are supplied to an auxiliary information comparing part 55. The output of the A/D converter 17 is supplied to an LPC analyzing part 56 and its spectrum envelope and pitch period power are detected and supplied to the comparing part 55. When a distance between both the spectrum envelopes is long, the comparing part 55 judges the existence of double talk and aborts impulse response estimation, and when the distance is short and both pitch periods are almost equal, an estimation speed is accelerated by increasing the correction step size of impulse response estimation. |
