Further tuning for AEC3 for initial echo suppression and handling of echo path changes

Further tuning for AEC3 for initial echo suppression 
and handling of echo path changes.

This CL add tuning for the AEC3 that
1) Improves the handling of the initial echo suppression
before the linear filter is reliable.
2) Improves the handling of echo path changes.

There are also minor bugfixes included.


BUG=webrtc:6018
Review-Url: https://codereview.webrtc.org/2717353002
Cr-Commit-Position: refs/heads/master@{#16873}
Committed: https://chromium.googlesource.com/external/webrtc/+/ebe7778ce9c28c2160dd6770cc58884a908382dd

[peah-webrtc, 2017-02-27 14:21:09]

Description was changed from

==========
Seems to work


Seems to work

BUG=
==========

to

==========
Further tuning for AEC3 to for initial echo suppression 
and handling of handle echo path changes

This CL add tuning for the AEC3 that
1) Improves the handling of the initial echo suppression
before the linear filter is reliable.
2) Improves the handling of echo path changes.

There are also minor bugfixes included.


BUG=webrtc:6018
==========

[peah-webrtc, 2017-02-27 14:26:02]

peah@webrtc.org changed reviewers:
+ henrik.lundin@webrtc.org

[peah-webrtc, 2017-02-27 14:26:03]

A new tuning CL for the AEC3.

PTAL

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
File webrtc/modules/audio_processing/aec3/echo_canceller3.cc (right):

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/aec3/echo_canceller3.cc:263:
data_dumper_->DumpWav("aec3_capture_analyze_input", capture->num_frames(),
The framelength was wrong here before.

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
File webrtc/modules/audio_processing/aec3/power_echo_model.cc (right):

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/aec3/power_echo_model.cc:31: constexpr float
kHInitial = 100.f;
An echo path can often have a gain of 20 dB so 10 dB was too small for an
initial gain.

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
File webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc (left):

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc:181: // Counts
the number of active render blocks that are in a row.
This counter was not correctly done.

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
File webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc (right):

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc:76: if
(active_render_blocks > kEchoPathChangeConvergenceBlocks) {
The filter-based echo path gain cannot be used until the filter is properly
converged. Hence use only the fallback echo power predictor until that is
achieved.

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc:157:
echo_path_gain_.fill(100.f);
Not important, but more appropriate.

https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc:184:
[kEchoPathGainHeadroom](float& a) { a *= kEchoPathGainHeadroom; });
Since the filter-based frequency response of the echo path is very course, a
headroom needs to be added.

[hlundin-webrtc, 2017-02-27 14:28:42]

On 2017/02/27 14:26:03, peah-webrtc wrote:
> A new tuning CL for the AEC3.
> 
> PTAL
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> File webrtc/modules/audio_processing/aec3/echo_canceller3.cc (right):
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> webrtc/modules/audio_processing/aec3/echo_canceller3.cc:263:
> data_dumper_->DumpWav("aec3_capture_analyze_input", capture->num_frames(),
> The framelength was wrong here before.
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> File webrtc/modules/audio_processing/aec3/power_echo_model.cc (right):
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> webrtc/modules/audio_processing/aec3/power_echo_model.cc:31: constexpr float
> kHInitial = 100.f;
> An echo path can often have a gain of 20 dB so 10 dB was too small for an
> initial gain.
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> File webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc (left):
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc:181: // Counts
> the number of active render blocks that are in a row.
> This counter was not correctly done.
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> File webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc (right):
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc:76: if
> (active_render_blocks > kEchoPathChangeConvergenceBlocks) {
> The filter-based echo path gain cannot be used until the filter is properly
> converged. Hence use only the fallback echo power predictor until that is
> achieved.
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc:157:
> echo_path_gain_.fill(100.f);
> Not important, but more appropriate.
> 
>
https://codereview.webrtc.org/2717353002/diff/1/webrtc/modules/audio_processi...
> webrtc/modules/audio_processing/aec3/residual_echo_estimator.cc:184:
> [kEchoPathGainHeadroom](float& a) { a *= kEchoPathGainHeadroom; });
> Since the filter-based frequency response of the echo path is very course, a
> headroom needs to be added.

Read the commit headline again...

[hlundin-webrtc, 2017-02-27 14:41:05]

LGTM is you fix the commit message headline.

[peah-webrtc, 2017-02-27 14:52:50]

Description was changed from

==========
Further tuning for AEC3 to for initial echo suppression 
and handling of handle echo path changes

This CL add tuning for the AEC3 that
1) Improves the handling of the initial echo suppression
before the linear filter is reliable.
2) Improves the handling of echo path changes.

There are also minor bugfixes included.


BUG=webrtc:6018
==========

to

==========
Further tuning for AEC3 for initial echo suppression 
and handling of echo path changes

This CL add tuning for the AEC3 that
1) Improves the handling of the initial echo suppression
before the linear filter is reliable.
2) Improves the handling of echo path changes.

There are also minor bugfixes included.


BUG=webrtc:6018
==========

[peah-webrtc, 2017-02-27 14:53:09]

Description was changed from

==========
Further tuning for AEC3 for initial echo suppression 
and handling of echo path changes

This CL add tuning for the AEC3 that
1) Improves the handling of the initial echo suppression
before the linear filter is reliable.
2) Improves the handling of echo path changes.

There are also minor bugfixes included.


BUG=webrtc:6018
==========

to

==========
Further tuning for AEC3 for initial echo suppression 
and handling of echo path changes.

This CL add tuning for the AEC3 that
1) Improves the handling of the initial echo suppression
before the linear filter is reliable.
2) Improves the handling of echo path changes.

There are also minor bugfixes included.


BUG=webrtc:6018
==========

[peah-webrtc, 2017-02-27 14:53:38]

On 2017/02/27 14:53:09, peah-webrtc wrote:
> Description was changed from
> 
> ==========
> Further tuning for AEC3 for initial echo suppression 
> and handling of echo path changes
> 
> This CL add tuning for the AEC3 that
> 1) Improves the handling of the initial echo suppression
> before the linear filter is reliable.
> 2) Improves the handling of echo path changes.
> 
> There are also minor bugfixes included.
> 
> 
> BUG=webrtc:6018
> ==========
> 
> to
> 
> ==========
> Further tuning for AEC3 for initial echo suppression 
> and handling of echo path changes.
> 
> This CL add tuning for the AEC3 that
> 1) Improves the handling of the initial echo suppression
> before the linear filter is reliable.
> 2) Improves the handling of echo path changes.
> 
> There are also minor bugfixes included.
> 
> 
> BUG=webrtc:6018
> ==========

Thanks! Updated the headline

[peah-webrtc, 2017-02-27 14:54:07]

The CQ bit was checked by peah@webrtc.org

[peah-webrtc, 2017-02-27 14:54:07]

The CQ bit was checked by peah@webrtc.org

[commit-bot: I haz the power, 2017-02-27 14:54:12]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.webrtc.org/...

[commit-bot: I haz the power, 2017-02-27 15:29:22]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1488207247717790, "parent_rev":
"ede0759c041715646f70ce4eb1fc85d69c963d18", "commit_rev":
"ebe7778ce9c28c2160dd6770cc58884a908382dd"}

[commit-bot: I haz the power, 2017-02-27 15:29:25]

Description was changed from

==========
Further tuning for AEC3 for initial echo suppression 
and handling of echo path changes.

This CL add tuning for the AEC3 that
1) Improves the handling of the initial echo suppression
before the linear filter is reliable.
2) Improves the handling of echo path changes.

There are also minor bugfixes included.


BUG=webrtc:6018
==========

to

==========
Further tuning for AEC3 for initial echo suppression 
and handling of echo path changes.

This CL add tuning for the AEC3 that
1) Improves the handling of the initial echo suppression
before the linear filter is reliable.
2) Improves the handling of echo path changes.

There are also minor bugfixes included.


BUG=webrtc:6018

Review-Url: https://codereview.webrtc.org/2717353002
Cr-Commit-Position: refs/heads/master@{#16873}
Committed:
https://chromium.googlesource.com/external/webrtc/+/ebe7778ce9c28c2160dd6770c...
==========

[commit-bot: I haz the power, 2017-02-27 15:29:26]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/external/webrtc/+/ebe7778ce9c28c2160dd6770c...