VoIP Voice and Fax Signal Processing

6

ECHO CANCELLATION

In telephonic voice conversation, echo is the return of a person's speech with delay, with reduced or modified sound level, and with a certain amount of distortions. In voice communication, echoes are created as acoustic and electrical. Acoustic echo is generated through an acoustic medium of a speakerphone or hands-free phone functionality. Electrical echoes are created in the two-to-four-wire telephone conversion hardware hybrids. Electrical echo is also called line echo.

In public switched telephone network (PSTN) local service, analog private branch exchange (PBX) systems, and VoIP, echoes are created inside the telephone and at two-to-four-wire telephone interface circuits. Echo in isolation is not a troubling signal in PSTN or VoIP. The increase in end-to-end delay makes echo perceivable and more annoying. In VoIP, end-to-end delay is caused by packet-based transmission. End-to-end delays also increase with higher packetization intervals and IP impediments to voice packets. The delays in PSTN voice calls are low even for local long-distance calls. Lower delay helps to tolerate echo as a sidetone. Sidetone is an attenuated version of the microphone signal fed back to the speaker. Sidetone returns immediately within 1-ms [ITU-T-G.107 (2005)] delay and for delays above 2 to 4 ms, a sidetone may be perceived as echo. In inter-regional PSTN calls, delays are in the order of 100 to 300 ms [ITU-T-G.114 (2003a)], and it becomes mandatory to incorporate echo cancellers ...