摘要 |
PROBLEM TO BE SOLVED: To estimate background noise in a voice signal encoder even when a user utters while monitoring with a speaker. SOLUTION: A silence detector 10 discriminates input audio signals for every 10 ms, which is a processing frame of an ITU-T recommendation G.729, to detect a silence frame. A which noise generator 11 generates small amplitude white noise signals. When a silence frame is detected, white noise signals generated by the generator 11 are supplied to an adder 12 for the frame, added to the input audio signals and supplied to a voice encoder 13. Then, the coder 13 performs initialization processes such as background noise estimation and the like and a silence compression process, and transmits the processed signals to a transmitting section 6. |