|
|
|
Because the ECAN is actually modeling the response of the tail circuit mathematically, it starts each conversation with no knowledge of which part of the signal is legitimate speech and which part is echo, so it has to build its model. This usually takes a few seconds. After that, the ECAN’s ability to discern and remove echo reaches its operating state for that call, and most echoes fade.
If a signal included multiple reverberations at different delays, like ripples in a pond, the ECAN would see perhaps the first two, but miss the one that arrived after 12 ms had passed. Typically, the energy level of each successive reverberation is reduced, so the resulting echo would potentially be quite soft anyway.
Distortion can also exceed the ability of ECANs operation. If the echo itself is so distorted that it does not truly match the pattern seen as the signal was sent into the tail circuit, the ECAN will not be able to recognize it as echo.
| Copyright © 2012 CA. All rights reserved. |
|