Talk:Echo suppression and cancellation

Nonsensical statement about wideband codecs
The page currently says,
 * On embedded platforms, it is not unusual to find a Wideband CODEC (such as AMR-WB / G.722) incorporated in place of an AEC. This said, many (embedded) VoIP solutions do have a fully functional AEC.

This makes no sense. The use of a wideband codec doesn't in any way reduce echo. To use a wideband codec "in place of" an echo canceller doesn't mean anything; the two have totally different and unrelated functions. I'll probably remove the statement about wideband codecs, which is the only nonsensical part. --Bigpeteb (talk) 13:54, 10 April 2014 (UTC)


 * No longer in article. ~Kvng (talk) 15:53, 18 August 2021 (UTC)

Acoustic echo cancellation
Acoustic echo cancellation redirects here but the Operation section describes only electronic echo (caused by hybrid imperfection) cancellation. ~Kvng (talk) 16:24, 3 July 2019 (UTC)


 * How so? That section says, "The filter is essentially a model of speaker, microphone and the room's acoustical attributes. Echo cancellers must be adaptive because the characteristics of the near-end's speaker and microphone are generally not known in advance. The acoustical attributes of the near-end's room are also not generally known in advance, and may change (e.g., if the microphone is moved relative to the speaker, or if individuals walk around the room causing changes in the acoustic reflections)." All of that describes acoustic echo. --Bigpeteb (talk) 22:57, 16 July 2019 (UTC)


 * In the telephone echo canceler, the adaptive filter cancels near end echo from all causes. Mostly, it is due to the hybrid transformer, but if there is anything causing echo out at the phone, it would cancel that too.  If the acoustic environment at the phone was changing, the adaptive filter would track that as long as it continued to adapt. We might add words to that effect.  In the original AT&T implementations, there was a filter adapted to each long distance trunk.  But the adapting algorithm was in shared equipment.  All the adaption took place in the first few seconds of conversation.  It would not track a changing echo as might be caused by the user changing his position relative to the speaker phone. But, now, that would not be a limitation. Constant314 (talk) 23:29, 16 July 2019 (UTC)


 * I guess I was referring more specifically to the diagram and caption in the Operation section. It would be nice to have a diagram that included the room. ~Kvng (talk) 14:33, 22 July 2019 (UTC)

Recent (Jul 7, 2023) changes
I disagree with your revert. I agree my changes involve jargon but are correct. The previous was too vague. Perhaps we can find something in between. Specifically, Constant314 (talk) 00:06, 9 July 2023 (UTC)
 * 1. The phrase "nature of the filter" is too vague. That could include, high-pass, low-pass, FIR, IIR, ladder, lattice, all sorts of topology questions. In practice, all of that is fixed and the adaption only changes the coefficients.  Would parameters be a better choice?
 * 2. "Near end signal" includes both near-end echo and near-end speech. There is no change chance of filtering the far-end signal into a replica of the near-end speech and even if that was possible, it would be a bad idea because you don't want to cancel the near-end speech.
 * Yours is the first occurrence of coefficients in the article. We could try to improve this by adding a link to Coefficient but that seems to send readers deeper into jargon soup. Coefficients directly determine the nature of the filter so I don't think this is too vague. If there needs to be a compromise, I might suggest response characteristics of the filter.
 * The filter is applied to the near-end signal and produces a signal that resembles the echo and this is subtracted from the signal received from the far end. So, on closer inspection, I don't think either of these descriptions is quite right. ~Kvng (talk) 01:05, 9 July 2023 (UTC)
 * response characteristics of the filter works for me. I will make that change. Constant314 (talk) 02:19, 9 July 2023 (UTC)
 * The filter is applied to the incoming signal from the farend. The output of the filter is subtracted from the near end signal.  The near-end signal isn't filtered (at least at the near-end). Constant314 (talk) 02:33, 9 July 2023 (UTC)
 * Thanks, your recent changes address my immediate concerns. The diagram in this article is my reference point and shows an echo generated by sound bouncing off the wall. When a telephone line is added there is some flexibility as to where the filter lives. The diagram in our article is a different application of the same principle – cancelling an electronic echo in the telephone transmission system not an acoustic echo in a room. It is not clear where the echo comes from in our diagram. ~Kvng (talk) 17:05, 11 July 2023 (UTC)
 * The caption refers to "filtered speech from the farend". That is the echo, although it is far from obvious.  I will try to explain it better. Constant314 (talk) 22:38, 11 July 2023 (UTC)