Adding second layer of the echo canceller 3 functionality.

webrtc/modules/audio_processing/aec3/render_delay_controller.h

+/*
+ *  Copyright (c) 2016 The WebRTC project authors. All Rights Reserved.

[hlundin-webrtc, 2017/01/18 13:08:50]

2017

[peah-webrtc, 2017/01/19 15:33:07]

Done.

+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef WEBRTC_MODULES_AUDIO_PROCESSING_AEC3_RENDER_DELAY_CONTROLLER_H_
+#define WEBRTC_MODULES_AUDIO_PROCESSING_AEC3_RENDER_DELAY_CONTROLLER_H_
+
+#include "webrtc/base/array_view.h"
+#include "webrtc/base/optional.h"
+#include "webrtc/modules/audio_processing/aec3/render_delay_buffer.h"
+#include "webrtc/modules/audio_processing/logging/apm_data_dumper.h"
+
+namespace webrtc {
+
+// Class for aligning the render and capture signal using an RenderDelayBuffer.
+class RenderDelayController {
+ public:
+  static RenderDelayController* Create(
+      ApmDataDumper* data_dumper,
+      int sample_rate_hz,
+      const RenderDelayBuffer& render_delay_buffer);
+  virtual ~RenderDelayController() = default;
+
+  // Aligns the render buffer content with the capture signal.
+  virtual size_t SelectDelay(rtc::ArrayView<const float> capture) = 0;

[hlundin-webrtc, 2017/01/18 13:08:50]

Select sounds a bit like the caller can select a delay through this call. Maybe
change it to CalculateDelay or EstimateDelay? Or GetAlignment[Delay]?

[peah-webrtc, 2017/01/19 15:33:07]

I agree. I changed it to GetDelay. I like that better than CalculateDelay and
EstimateDelay as the class is not called something with estimator. Furthermore,
I prefer that to GetAlignment as the class does not perform the alignment but
rather controls the delay in the render delay buffer.

WDYT?

[hlundin-webrtc, 2017/01/20 09:31:44]

Acknowledged.

[aleloi, 2017/01/20 14:45:37]

GetDelay sounds fine.

[peah-webrtc, 2017/01/23 14:16:39]

Acknowledged.

[peah-webrtc, 2017/01/23 14:16:39]

Acknowledged.

+
+  // Analyses the render signal and returns false if there is a buffer overrun.
+  virtual bool AnalyzeRender(rtc::ArrayView<const float> render) = 0;
+
+  // Returns an approximate value for the headroom in the buffer alignment
+  // expressed in samples.

[hlundin-webrtc, 2017/01/18 13:08:50]

When does it return empty?

[peah-webrtc, 2017/01/19 15:33:07]

What it does is to return the alignment headroom that would be achieved if the
delay provided by the controller is applied to the render delay buffer. That it
only does, however, if the controller has confidence in the delay that is
reported, hence the possible empty return.

For instance, consider the case where the delay controller at first finds good
correlation between render and capture and from that chooses a render buffer
delay, but then for a long while does not find any correlation. In that case,
the chosen render buffer delay should remain the same, but it cannot promise any
reliable alignment headroom.

Maybe we should change the name for the AlignmentHeadroom method as that is
misleading as the controller does not align the signals. This rather the
headroom that is achieved if the provided delay is applied to the render buffer.

WDYT?

[hlundin-webrtc, 2017/01/20 09:31:44]

I think some of this explanation should go into the code.

[peah-webrtc, 2017/01/23 14:16:39]

I changed the name of the method to achieve part of comment, and let the return
of the optional value replace another part of the comment.
PTAL

+  virtual rtc::Optional<size_t> AlignmentHeadroom() const = 0;
+};
+}  // namespace webrtc
+
+#endif  // WEBRTC_MODULES_AUDIO_PROCESSING_AEC3_RENDER_DELAY_CONTROLLER_H_