The estimation on how well the linear filter in the AEC3 is performing
is done through an estimation of the ERLE. That estimation is then
used for knowing how much the suppressor needs to react in order to
cancel all the echoes.
In the current code, the ERLE is quite conservative during farend
inactivity and it is common that it goes to a minimum value during
those periods. Under highly varying conditions, that is probably the
right approach. However, in other scenarios where conditions does not
change that fast there is a loss in transparency that could be avoided
by means of a different ERLE estimation.
In the current CL, the ERLE estimation has been changed in the
following way:
- During farend activity the ERLE is estimated through a 1st order AR
smoother. This smoother goes faster toward lower ERLE values than to
larger ones in order to avoid overestimation of this
value. Furthermore, during the beginning of the farend burst, an
estimation of the ERLE is done that aim to represent the performance
of the linear filter during onsets. Under highly variant environments,
those quantities, the ERLE during onsets and the one computed during
the whole farend duration, would differ a lot. If the environment is
more stationary, those quantities would be much more similar.
- During nearend activity the ERLE estimation is decreased toward a
value of the ERLE during onsets.
Bug: webrtc:9040
Change-Id: Ieab86370a4333d2d0cd7041047d29651de4f6827
Reviewed-on: https://webrtc-review.googlesource.com/62342
Commit-Queue: Jesus de Vicente Pena <devicentepena@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22568}
This CL adds robustifications for avoiding that the headset mode
is triggered for reverberant or weak echo paths.
Bug: webrtc:9047,chromium:824111,webrtc:8314,webrtc:8671,webrtc:5201,webrtc:5919
Change-Id: Ib111e617f765377c021a5b633cf13a7917fe62a6
Reviewed-on: https://webrtc-review.googlesource.com/64002
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22557}
This CL robustifies the echo removal behavior when headsets are used.
In particular it:
-Introduces a secondary, more refined alignment when no alignment can
be found using the delay estimator.
-Changes decision logic for when to use the linear filter output.
-Changes the decision logic for when to be transparent.
-Changes the way that the transparent mode works.
-Makes the nonlinear mode less aggressive.
-Removes the detector for non-audible echoes.
-Makes the attenuation when there are signals with strong narrowband
characteristics more mild in scenarios with low render.
Furthermore the CL:
-Removes the input of external echo leakage information.
Bug: webrtc:9047,chromium:824111,webrtc:8314,webrtc:8671,webrtc:5201,webrtc:5919
Change-Id: Ied1fe0c0a35d3c31b47606ed2db319a73644d406
Reviewed-on: https://webrtc-review.googlesource.com/60866
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22548}
This CL significantly improves the response time
of the AEC3 delay estimator to audio buffer issues.
The CL adds ensures that the delay estimator
correlators reacts to buffer issues from the
zero state which is much faster than if it has already
achieved a state matching a previous alignment.
The CL has been extensively tested on offline
recordings.
Bug: webrtc:9023, chromium:822245
Change-Id: Ic149b9429e592d4c3535eb8432582f435a1b4745
Reviewed-on: https://webrtc-review.googlesource.com/62081
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Ivo Creusen <ivoc@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22461}
This CL ensures a smooth transition from the parameters used during
the startup phase in the call to the parameters used in the rest of the
call. This is achieved by slowly transitioning between the parameter
sets via interpolation.
Bug: chromium:819240,webrtc:8983
Change-Id: Ifbac4b93fc6ad6efc441f41fb88ef09e8ee3d669
Reviewed-on: https://webrtc-review.googlesource.com/60360
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22451}
This CL corrects the bug that only looked at narrowband
render signals above 900 Hz and only assumed that the
influence of such lasted for 6 blocks, which resulted
in filter divergence and echo leakage.
Bug: webrtc:9008,chromium:821670
Change-Id: I9b2635d24b260e9d9a8c5c088ab663e03fb93c42
Reviewed-on: https://webrtc-review.googlesource.com/61800
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Ivo Creusen <ivoc@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22434}
So that we can avoid dependency cycles.
Bug: none
Change-Id: I821d9f1319dff01403d6e4e310cbb2d4b2b125e8
Reviewed-on: https://webrtc-review.googlesource.com/60500
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Commit-Queue: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22328}
This CL introduces a different rampup behavir for the call startup and after resets
that may occur due to delay changes, clock-drift and audio path glitches.
Bug: chromium:819111, webrtc:8979
Change-Id: Ied1d7896be7f0c69aa6deb61475117021ca6ab09
Reviewed-on: https://webrtc-review.googlesource.com/60002
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22312}
Note: estimation is turned OFF if config_.ep_strength.default_len
is set >= 0 (in this case config_.ep_strength.default_len defines a
constant echo decay factor), and hence turned ON if < 0. In case the
echo tail estimation is turned ON, -config_.ep_strength.default_len is
the starting point for the estimator.
The estimation is done in two passes; first we go through all "sections"
(corresponding to chunks of length kFftLengthBy2) of the filter impulse
response to determine which sections correspond to a "stable" decay",
and then the second pass we go through each stable decay section and
estimate the decay. The actual decay estimation is based on linear
regression of the log magnitude of the squared impulse response.
A bunch of sanity checks are also performed continuously to avoid
estimation error during e.g., filter adaptation.
Bug: webrtc:8924
Change-Id: I686ce3f3e8b6b472348f8d6e01fb44c31e25145d
Reviewed-on: https://webrtc-review.googlesource.com/48440
Commit-Queue: Christian Schuldt <cschuldt@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22247}
The AEC3 factory is now part of the WebRTC API.
Bug: webrtc:8844
Change-Id: If6f419b4ca0354e2d346c0e6474086e456ba747e
Reviewed-on: https://webrtc-review.googlesource.com/57141
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Reviewed-by: Alex Loiko <aleloi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22204}
This CL moves the AEC3 API call skew estimator into a separate file.
This has the advantage that it can more easily be tested.
The CL also simplifies the code and adds unittests.
Bug: webrtc:8671
Change-Id: I19bc31ca5666cdc87a1ed14770ef20ead1b5b80d
Reviewed-on: https://webrtc-review.googlesource.com/55860
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22144}
This CL adds functionality for passing the information about the
estimated delay to the echo remover in AEC3.
The CL also adds information about how long ago the delay changed,
and how long ago the delay estimate was updated.
Bug: webrtc:8671
Change-Id: If274ffe0465eb550f3e186d0599c6dc6fef7f5e8
Reviewed-on: https://webrtc-review.googlesource.com/55261
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22137}
This CL adds functionality to allow removal of any echo occurring
before the render and capture signals have been properly aligned.
The functionality is added in such a manner that the transparency
to nearend is maintained as much as possible.
Bug: webrtc:8883
Change-Id: I813cbbc4c48822e7dffcd9ab6233be4c222089de
Reviewed-on: https://webrtc-review.googlesource.com/49941
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22107}
Avoid including audio_processing.h from within AEC3.
Bug: webrtc:8844
Change-Id: I02c475c2fb84e2c24eac86baac3c7edaa08bebc0
Reviewed-on: https://webrtc-review.googlesource.com/53065
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22029}
This CL resets the AEC3 realignment functionality when a significant
and persistent skew in the number of render and capture API calls is
detected.
Bug: chromium:811658,webrtc:8879
Change-Id: Ib5c727b38f427da2a7d25eac7c939a17bdaabe74
Reviewed-on: https://webrtc-review.googlesource.com/52260
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21997}
This CL adds robustness in terms of echo removal and faster recovery
in order to regain echo canceller transparency after echo path changes.
The CL does:
-Improve the adaptation rate of the linear filter.
-Increase the look-window used before the linear filter has adapted.
-Decrease the effects of missed detection of residual echo.
-Increase the safety margin before allowing the suppressor gain to
increase.
Bug: chromium:804873,webrtc:8788
Change-Id: I28eedc4c8d0a4f0bc7b79c02d6d59bf00fddd566
Reviewed-on: https://webrtc-review.googlesource.com/48721
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21917}
The faster AEC3 alignment introduced recently may in
cases cause the alignment (and the AEC3) to repeatedly
reset. This CL avoids these resets by handling buffer
issues (which are triggering the resets) separately
during the initial coarse alignment phase.
Change-Id: Idf5e2ffda2591906da8060d03ec8ca73cdaedf53
Bug: webrtc:8798,chromium:805815
Reviewed-on: https://webrtc-review.googlesource.com/43480
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21758}
This CL increases the speech of the initial alignment in AEC3 by
loosening the requirements on the accuracy of the initial estimates.
Bug: webrtc:8784, chromium:804270
Change-Id: I86e2d97830843524090a1cf877965739f66dc058
Reviewed-on: https://webrtc-review.googlesource.com/40660
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21728}
This CL generalizes the hysteresis behavior on the AEC3 delay estimator
to be two-sided and easier to configure.
Bug: webrtc:8671
Change-Id: Ife21c1511416e32eb3618c81178deefe332ac1e8
Reviewed-on: https://webrtc-review.googlesource.com/39267
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21604}
This CL adds an nonwindowed spectrum of the linear filter error
to use in the NLP computation.
Bug: webrtc:8661
Change-Id: I45bc9bb3eb8eeac0c5d6adb414638eb12b635a27
Reviewed-on: https://webrtc-review.googlesource.com/38701
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21583}
This CL corrects the way that the estimated filter delay is used in
AEC3. In particular
-It uses the filter delay to choose the correct render block in AecState
-It changes the code to reflect that the filter delay is always computed
-It removes part of the code that formerly relied on the filter delay
being an Optional.
Bug: webrtc:8671
Change-Id: I58135a5c174b404707e19a41c3617c09831e871d
Reviewed-on: https://webrtc-review.googlesource.com/35221
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21557}
Changing window type which improves the filter accuracy
at the cost of a slight reduction in convergence time.
Bug: webrtc:8661
Change-Id: Id0e5c66ec179f94471cbca0a2b8d1b94d8146ca6
Reviewed-on: https://webrtc-review.googlesource.com/34501
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21347}
This CL adds windowing of the error signal in echo canceller 3 to
avoid issues with spectral leakage affecting the quality of
the filter estimate.
Bug: webrtc:8661
Change-Id: I3e583f80fe02d7bba387a906bf44fbe7b89a2a6f
Reviewed-on: https://webrtc-review.googlesource.com/34188
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21328}
This CL adjusts the filter adaptation behavior to better handle
reverberant environments and environments with poor SNR.
It furthermore updates the unittests to handle the reduced adaptation
speed.
Bug: webrtc:8661
Change-Id: I5f1b5a4a34b333bd6c643ed3727899d0838dbf90
Reviewed-on: https://webrtc-review.googlesource.com/34184
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#21323}
