The ERLE is used to estimate residual echo for echo suppression. The
ERLE is reduced during far-end offset to avoid echo leakage. When there
is a strong near-end present this can cause unnecessary transparency loss.
This change adds an ERLE estimation that does not compensate for onsets and
uses it for residual echo estimation when the suppressor considers the near-end to be dominant.
Bug: webrtc:12686
Change-Id: Ida78eeacf1f95c6e62403f86ba3f2ff055898a84
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215323
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33786}
Erle Uncertainty changes the residual echo computation during saturated
echo. However, the case of saturated echo is already handled by the
residual echo estimator causing the ErleUncertainty to be a no-op.
The change has been tested for bit-exactness.
Bug: webrtc:8671
Change-Id: I779ba67f99f29d4475a0465d05da03d42d50e075
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215072
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33719}
After the refined filter has been determined to perform better than
the coarse filter, and the coefficients of the coarse filters are
overwritten by the ones from the refined filter, at least 100 ms have
to pass before the adaptation of the refined filter is allowed to speed
up due to good coarse filter performance.
This change solves the vicious circle described in webrtc:12265, where
the coarse and refined filters can diverge over time.
This feature can be disabled remotely via a kill-switch. When disabled
the AEC output is bit-exact to before the change.
Bug: webrtc:12265,chromium:1155477
Change-Id: Iacd6e325e987dd8a475bb3e8163fee714c65b20a
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/196501
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32801}
This change makes the transparent mode classifier consider filter
convergence also for microphone signals with very low volume. This is
in order to prevent entering transparent mode when there is low, but
audible, echo.
Furthermore, the suppression gain during transparent mode is restored
to avoid leaks when the echo is too low to be reliably detected by
filter convergence.
Bug: webrtc:10232, chromium:1140452
Change-Id: Idd4f40c4aee7c20baa444afaa5ec604eb65bcfd0
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/189786
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32461}
This CL adds delay headroom when an external delay estimator is used.
Tested: audioproc_f --aec=1 is bitexact on a large number of aecdumps
Bug: b/158455753
Change-Id: I56de44e841bb8162e302181c6c386ad7fbb00dee
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/186703
Commit-Queue: Sam Zackrisson <saza@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32323}
This change introduces a new Hidden Markov Model based classifier for
AEC3's 'transparent mode'. Transparent mode is used with
headsets/headphones where the speaker signal does not leak into the
microphone signal.
The current classifier suffers from two problems:
1. It sometimes takes a long time to enter transparent mode.
2. Sometimes transparent mode is left (and it once again takes a long
time to re-enter).
Both problems have a severe effect on AEC transparency.
The new classifier enters transparent mode quicker and is less likely
to exit transparent mode when there is no echo. This improves the
audio experience when using headset/headphones.
Another (minor) benefit of this change is that when transparent mode
is disabled no classifier is run (or even created) saving some memory
and CPU cycles.
Bug: webrtc:10232
Change-Id: I509af0e22b59463aeaead53c78c35be1e97fe8c3
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/184500
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32182}
This CL adds a number of kill-switches to the AEC3 code to be used as
safe fallbacks to increase AEC transparency.
The changes have been shown to be bitexact for a test dataset.
Bug: webrtc:11475,chromium:1066836
Change-Id: Ibebcbbfbbd958cb6fcc6993247e3030fa65b582c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/172600
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#30964}
This CL renames the shadow filter in AEC3 to have the more accurate name
coarse filter.
The CL consists of 3 main initial patch sets, designed to simplify
the review:
1) Replaces "shadow" with "coarse" and adds a fall-back functionality
to support the old filter naming.
2) Renames the files according to the new naming.
3) Performs a "git cl format"
Bug: webrtc:8671
Change-Id: I28d6041d0d34e85f8f8048d004b44a1a5f07bb07
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/170981
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#30846}
This CL renames the main filter in AEC3 to have the more accurate name
refined filter.
The CL consists of 3 main initial patch sets, designed to simplify
the review:
1) Replaces "main" with "refined" and adds a fall-back functionality
to support the old filter naming.
2) Renames the files according to the new naming.
3) Performs a "git cl format"
Bug: webrtc:8671
Change-Id: Ifd0aab34e291736a2250e0986348404618630b1d
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/170825
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#30843}
The estimator will simply compute the worst value of all combinations
of render and capture signal.
This has the drawback that low-volume or silent render channels may
severely misestimate the ERL.
The changes have been shown to be bitexact over a large dataset.
Bug: webrtc:10913
Change-Id: Id53c3ab81646ac0fab303edafc5e38892d285d8e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/157308
Commit-Queue: Sam Zackrisson <saza@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29542}
- RenderBuffer::Spectrum() loses its channel argument, allowing for
greater flexibility in passing the multi-channel spectrum data into
functions.
- The FFT spectra lengths are made compile-time constant, rendering
some DCHECKs obsolete.
Bug: webrtc:10913
Change-Id: Ied0c50cf72d974cfef7279fd2b9c572d049b8b16
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/157104
Commit-Queue: Sam Zackrisson <saza@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29528}
This CL corrects the EchoAudibility and StationarityEstimator
code to work properly with multiple channels.
It also changes the naming of the FilterDelayBlocks() method
to better reflect what it does.
The changes have been verified to be bitexact over a large number
of recordings.
Bug: webrtc:10913
Change-Id: I070b531efcdff4c33f70fd5b37fbb556dcebe5b4
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156565
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29482}
This CL adds support for multiple channels in the reverb
modelling. As a side effect, it also partly adds multi-channel
supports for the sections of the code.
Beyond adding the multi-channel support, a bug is fixed as part of
this CL. Since the bug fix affects the bitexactness, as a safety
precaution the CL includes the ability to override the bugfix.
Apart from the contributions from the bugfix, the changes have
been verified to be bitexact for a large set of mono recordings.
Bug: webrtc:10913
Change-Id: I1f307b532be85ef4182f8db41384f44d40a25219
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156382
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29456}
This CL passes the spectral power estimates for all channels into
the AecState.
Bug: webrtc:10913
Change-Id: Ie3b5c443be0c63f205e23ed2bfea06d9c447eb39
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/156165
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29417}
This CL removes the redundant class in preparation
for adding multichannel functionality to the
reverb computation.
The changes are bitexact.
Bug: webrtc:10913
Change-Id: I284665f7143cb5e1c79bfa573638fdff5f2411c9
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/155960
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29414}
Multi-channel behaviors introduced in this CL:
- All filters are analyzed independently. The filtering is considered
consistent if any filter is consistent.
- The filter echo path gain used to detect saturation is maxed across
capture channels.
- The filter delay is taken to be the minimum of all filters:
Any module that looks in the render data starting from the filter
delay will iterate over all render audio present in any channel.
- The FilterAnalyzer will consider a render block to be active if any
render channel has activity.
The changes in the CL has been shown to be bitexact on a
large set of mono recordings.
Bug: webrtc:10913
Change-Id: I1e360cd7136ee82d1f6e0f8a1459806e83f4426d
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/155363
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29408}
This CL adds support for multichannel in the residual echo
estimator code. It also adds placeholder functionality in
the surrounding code to ensure that the residual echo
estimator receives the require inputs.
The changes in the CL has been shown to be bitexact on a
large set of mono recordings.
Bug: webrtc:10913
Change-Id: I726128ca928648b1dcf36c5f479eb243f3ff3f96
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/155361
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29400}
All dependent modules are hardcoded to do their regular mono processing on the first channel.
This _almost_ makes RenderBuffer multi-channel: FftData is still only mono.
Bug: webrtc:10913
Change-Id: Id5cc34dbabfe59e1cc72a9675dc7979794e870ed
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/151139
Commit-Queue: Sam Zackrisson <saza@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29074}
This is a reland of a66395e72f
Original change's description:
> Reland "Add core multi-channel pipeline in AEC3 This CL adds basic the basic pipeline to support multi-channel processing in AEC3."
>
> This is a reland of f3a197e553
>
> Original change's description:
> > Add core multi-channel pipeline in AEC3
> > This CL adds basic the basic pipeline to support multi-channel
> > processing in AEC3.
> >
> > Apart from that, it removes the 8 kHz processing support in several
> > places of the AEC3 code.
> >
> > Bug: webrtc:10913
> > Change-Id: If5b75fa325ed0071deea94a7546cb4a7adf22137
> > Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/150332
> > Commit-Queue: Per Åhgren <peah@webrtc.org>
> > Reviewed-by: Sam Zackrisson <saza@webrtc.org>
> > Cr-Commit-Position: refs/heads/master@{#29017}
>
> Bug: webrtc:10913
> Change-Id: Ifc4b13bd994cfd22dca8f8755fa5700617cc379d
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/151124
> Reviewed-by: Sam Zackrisson <saza@webrtc.org>
> Commit-Queue: Per Åhgren <peah@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#29034}
Bug: webrtc:10913
Change-Id: Id8da5666df8c86f290c73ad5dc9958199f1a7ebe
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/151127
Commit-Queue: Sam Zackrisson <saza@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29042}
This is a reland of f3a197e553
Original change's description:
> Add core multi-channel pipeline in AEC3
> This CL adds basic the basic pipeline to support multi-channel
> processing in AEC3.
>
> Apart from that, it removes the 8 kHz processing support in several
> places of the AEC3 code.
>
> Bug: webrtc:10913
> Change-Id: If5b75fa325ed0071deea94a7546cb4a7adf22137
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/150332
> Commit-Queue: Per Åhgren <peah@webrtc.org>
> Reviewed-by: Sam Zackrisson <saza@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#29017}
Bug: webrtc:10913
Change-Id: Ifc4b13bd994cfd22dca8f8755fa5700617cc379d
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/151124
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29034}
This reverts commit f3a197e553.
Reason for revert: Speculative revert, as this may'be broken some build bots
Original change's description:
> Add core multi-channel pipeline in AEC3
> This CL adds basic the basic pipeline to support multi-channel
> processing in AEC3.
>
> Apart from that, it removes the 8 kHz processing support in several
> places of the AEC3 code.
>
> Bug: webrtc:10913
> Change-Id: If5b75fa325ed0071deea94a7546cb4a7adf22137
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/150332
> Commit-Queue: Per Åhgren <peah@webrtc.org>
> Reviewed-by: Sam Zackrisson <saza@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#29017}
TBR=saza@webrtc.org,peah@webrtc.org
Change-Id: I877d2993b9ccf024bd1d57bca1513c3e24d0bed3
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: webrtc:10913
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/150940
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29022}
This CL adds basic the basic pipeline to support multi-channel
processing in AEC3.
Apart from that, it removes the 8 kHz processing support in several
places of the AEC3 code.
Bug: webrtc:10913
Change-Id: If5b75fa325ed0071deea94a7546cb4a7adf22137
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/150332
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29017}
This CL removes a long unused fallback behavior for the reverb
computation.
Bug: webrtc:8671
Change-Id: I4b57795a9bb33769237858f40392881ee235653e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/148520
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#28802}
The configuration parameter filter.use_linear_filter can be used to
disable the linear filtering. Disabling the linear filter is equivalent
to runing in non-linear mode.
Bug: b/130016532
Change-Id: I8ffdf474822888b9915444bba6cc1c25ec1efe5a
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/132552
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#27566}
This change reduces the risk of echo due to noise in the headroom
of the linear filter.
Changes:
- Use shorter delay headroom
- Delay headroom is specified in samples (not blocks)
- No hysteresis limit when delay is reduced
Bug: chromium:119942,webrtc:10341
Change-Id: I708e80e26d541dff8ca04b6da2d346a1d59cbfcb
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/126420
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#27126}
In this CL the analysis of the impulse response that is done in the FilterAnalyzed class is changed in order to reduce its complexity. Instead of analyzing the whole impulse response in each Update call a smaller region is analyzed. That region is changed at each Update call which implies that several calls are needed in order to analyze the complete impulse response.
Bug: webrtc:10032,chromium:909007
Change-Id: Ic58be34ba18485311c63e0fed9b6e892f9cb864c
Reviewed-on: https://webrtc-review.googlesource.com/c/111602
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Commit-Queue: Jesus de Vicente Pena <devicentepena@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#25817}
In this CL a more precise estimation of the Erle is introduced. This is done by creating different estimators that are specialized in different regions of the linear filter. An estimation of which regions were used for generating the current echo estimate is performed and used for selecting the right Erle estimator.
Bug: webrtc:9961
Change-Id: Iba6eb24596c067c3c66d40df590be379d3e1bb7b
Reviewed-on: https://webrtc-review.googlesource.com/c/109400
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Jesus de Vicente Pena <devicentepena@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#25707}
This CL is the result of running include-what-you-use tool on part
of the code base (audio target and dependencies) plus manual fixes.
bug: webrtc:8311
Change-Id: I277d281ce943c3ecc1bd45fd8d83055931743604
Reviewed-on: https://webrtc-review.googlesource.com/c/106280
Commit-Queue: Yves Gerey <yvesg@google.com>
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Patrik Höglund <phoglund@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#25311}
In this CL we change the signal that controls the updates of the ERLE estimator. Until now, the render signal was used which is not optimum for reverberant signals. In this CL, a reverberation has been added to the the render signal and this new signal has been used for controlling when to update the ERLE estimator.
Bug: webrtc:9873
Change-Id: I0ebea3fc208f97aa237af015ba543015d49ed978
Reviewed-on: https://webrtc-review.googlesource.com/c/105660
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Jesus de Vicente Pena <devicentepena@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#25285}
Removing code that has no audible effect.
Bug: webrtc:8671
Change-Id: Ibd7d0d19d760ae16b09285498c2ee09b42eb5968
Reviewed-on: https://webrtc-review.googlesource.com/c/106301
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#25250}
This CL introduces two related changes
1) It changes the way that the AEC3 determines whether the linear
filter is sufficiently good for its output to be used. The new scheme
achieves this much earlier than what was done in the legacy scheme.
2) It changes the way that saturated echo is and handled so that the
impact of the nearend speech is lower.
Bug: webrtc:9835,webrtc:9843,chromium:895435,chromium:895431
Change-Id: I0b493676886e2134205e9992bbe4badac7e414cc
Reviewed-on: https://webrtc-review.googlesource.com/c/104380
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#25208}
This CL introduces a major refactoring of AecState for the purpose of
simplifying further improvements to the logic in this code.
The changes have successfully been tested for bitexactness.
Bug: webrtc:8671
Change-Id: If98efde55a22c76b093089a11a0562daac7e16e6
Reviewed-on: https://webrtc-review.googlesource.com/c/102362
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#24996}
In this CL the ERLE estimator is reset after a pre-amplifier gain change is communicated to APM.
Bug: webrtc:9805
Change-Id: I040f344e4607e862240250f9478d06de0d58a096
Reviewed-on: https://webrtc-review.googlesource.com/103222
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Reviewed-by: Alex Loiko <aleloi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#24933}
This CL removes killswitches for code that has been properly tested in
experiments and is to be considered to be permanent.
The changes have been tested for bitexactness.
Bug: webrtc:8671
Change-Id: I0f9db16f377390d9dd3779096da91f3abc0fb4a5
Reviewed-on: https://webrtc-review.googlesource.com/102360
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#24877}
This CL adjusts the behavior of the AEC3 echo suppressor behavior
initially in the call, and when there has been delay changes. The
results is that short echo blips/bursts present in some such cases
no longer occur.
In particular this CL:
-Ensures that the suppressor back-off under stationary render
conditions does not occur until the linear filter has had the
ability to converge.
-Ensures that a previously converged filter behavior detection
is not sticky for stable and linear echo paths, which in turn
prevents echo leakage due to the more liberal echo suppressor
behavior applied on such platforms.
-Removes a bug that caused a random and jittery behavior for
the usage of the linear filter output initially in the calls
and after echo path changes
Bug: webrtc:9737, chromium:882396
Change-Id: Id2b46e366dc58ab8137f19ed59a2034c89ca3087
Reviewed-on: https://webrtc-review.googlesource.com/99063
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#24656}
