Add FixedGainController and move GainController2 in APM.
The FixedGainController (FGC) applies a fixed gain. It will also
control the limiter. The limiter will be landed over the next several
CLs.
The GainController2 is a 'private submodule' of APM. It will control
the new automatic gain controller (AGC). It controls the AGC through
Initialize() and ApplyConfig().
This CL contains
* build changes to make modules/audio_processing/agc2 an independent
target
* a new MutableFloatAudioFrame which is the audio interface between
AGC2 and APM
* move of the fixed gain application from GainController2 to
FixedGainController.
If you are a googler, there is more information in this doc:
https://docs.google.com/document/d/1RV2Doet3MZtUPAHVva61Vjo20iyd1bmmm3aR8znWpzo/edit#
Bug: webrtc:7949
Change-Id: Ief95cbbce83c3aafe54638fd2ab881c9fb8bdc3a
Reviewed-on: https://webrtc-review.googlesource.com/50440
Commit-Queue: Alex Loiko <aleloi@webrtc.org>
Reviewed-by: Oskar Sundbom <ossu@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#22046}diff --git a/modules/audio_processing/BUILD.gn b/modules/audio_processing/BUILD.gn
index ba0cf99..60605e9 100644
--- a/modules/audio_processing/BUILD.gn
+++ b/modules/audio_processing/BUILD.gn
@@ -120,8 +120,6 @@
"agc/loudness_histogram.h",
"agc/utility.cc",
"agc/utility.h",
- "agc2/gain_controller2.cc",
- "agc2/gain_controller2.h",
"audio_buffer.cc",
"audio_buffer.h",
"audio_processing_impl.cc",
@@ -151,6 +149,8 @@
"gain_control_for_experimental_agc.h",
"gain_control_impl.cc",
"gain_control_impl.h",
+ "gain_controller2.cc",
+ "gain_controller2.h",
"include/audio_processing.cc",
"include/audio_processing.h",
"include/config.cc",
@@ -215,6 +215,7 @@
":aec_core",
":aec_dump_interface",
":apm_logging",
+ ":audio_frame_view",
":audio_processing_c",
":audio_processing_statistics",
"..:module_api",
@@ -234,6 +235,7 @@
"../../system_wrappers:cpu_features_api",
"../../system_wrappers:field_trial_api",
"../../system_wrappers:metrics_api",
+ "agc2",
"vad",
]
@@ -287,6 +289,15 @@
]
}
+rtc_source_set("audio_frame_view") {
+ sources = [
+ "include/audio_frame_view.h",
+ ]
+ deps = [
+ "../../api:array_view",
+ ]
+}
+
rtc_source_set("aec_dump_interface") {
sources = [
"include/aec_dump.cc",
@@ -294,6 +305,7 @@
]
deps = [
+ ":audio_frame_view",
"../../api:array_view",
"../../rtc_base:rtc_base_approved",
]
@@ -585,6 +597,7 @@
"../../test:test_support",
"../audio_coding:neteq_input_audio_tools",
"aec_dump:mock_aec_dump_unittests",
+ "agc2:fixed_digital_unittests",
"test/conversational_speech:unittest",
"vad:vad_unittests",
"//testing/gtest",
@@ -653,7 +666,6 @@
"aec3/suppression_filter_unittest.cc",
"aec3/suppression_gain_unittest.cc",
"aec3/vector_math_unittest.cc",
- "agc2/gain_controller2_unittest.cc",
"audio_processing_impl_locking_unittest.cc",
"audio_processing_impl_unittest.cc",
"audio_processing_unittest.cc",
@@ -665,6 +677,7 @@
"echo_detector/moving_max_unittest.cc",
"echo_detector/normalized_covariance_estimator_unittest.cc",
"gain_control_unittest.cc",
+ "gain_controller2_unittest.cc",
"level_controller/level_controller_unittest.cc",
"level_estimator_unittest.cc",
"low_cut_filter_unittest.cc",
diff --git a/modules/audio_processing/aec_dump/aec_dump_impl.cc b/modules/audio_processing/aec_dump/aec_dump_impl.cc
index 594bf85..4deb192 100644
--- a/modules/audio_processing/aec_dump/aec_dump_impl.cc
+++ b/modules/audio_processing/aec_dump/aec_dump_impl.cc
@@ -94,11 +94,13 @@
worker_queue_->PostTask(std::unique_ptr<rtc::QueuedTask>(std::move(task)));
}
-void AecDumpImpl::AddCaptureStreamInput(const FloatAudioFrame& src) {
+void AecDumpImpl::AddCaptureStreamInput(
+ const AudioFrameView<const float>& src) {
capture_stream_info_.AddInput(src);
}
-void AecDumpImpl::AddCaptureStreamOutput(const FloatAudioFrame& src) {
+void AecDumpImpl::AddCaptureStreamOutput(
+ const AudioFrameView<const float>& src) {
capture_stream_info_.AddOutput(src);
}
@@ -134,7 +136,8 @@
worker_queue_->PostTask(std::unique_ptr<rtc::QueuedTask>(std::move(task)));
}
-void AecDumpImpl::WriteRenderStreamMessage(const FloatAudioFrame& src) {
+void AecDumpImpl::WriteRenderStreamMessage(
+ const AudioFrameView<const float>& src) {
auto task = CreateWriteToFileTask();
auto* event = task->GetEvent();
diff --git a/modules/audio_processing/aec_dump/aec_dump_impl.h b/modules/audio_processing/aec_dump/aec_dump_impl.h
index 5be876b..36d72e9 100644
--- a/modules/audio_processing/aec_dump/aec_dump_impl.h
+++ b/modules/audio_processing/aec_dump/aec_dump_impl.h
@@ -54,15 +54,16 @@
void WriteInitMessage(const InternalAPMStreamsConfig& api_format) override;
- void AddCaptureStreamInput(const FloatAudioFrame& src) override;
- void AddCaptureStreamOutput(const FloatAudioFrame& src) override;
+ void AddCaptureStreamInput(const AudioFrameView<const float>& src) override;
+ void AddCaptureStreamOutput(const AudioFrameView<const float>& src) override;
void AddCaptureStreamInput(const AudioFrame& frame) override;
void AddCaptureStreamOutput(const AudioFrame& frame) override;
void AddAudioProcessingState(const AudioProcessingState& state) override;
void WriteCaptureStreamMessage() override;
void WriteRenderStreamMessage(const AudioFrame& frame) override;
- void WriteRenderStreamMessage(const FloatAudioFrame& src) override;
+ void WriteRenderStreamMessage(
+ const AudioFrameView<const float>& src) override;
void WriteConfig(const InternalAPMConfig& config) override;
diff --git a/modules/audio_processing/aec_dump/capture_stream_info.cc b/modules/audio_processing/aec_dump/capture_stream_info.cc
index e3284d8..dd48fd4 100644
--- a/modules/audio_processing/aec_dump/capture_stream_info.cc
+++ b/modules/audio_processing/aec_dump/capture_stream_info.cc
@@ -19,7 +19,7 @@
CaptureStreamInfo::~CaptureStreamInfo() = default;
-void CaptureStreamInfo::AddInput(const FloatAudioFrame& src) {
+void CaptureStreamInfo::AddInput(const AudioFrameView<const float>& src) {
RTC_DCHECK(task_);
auto* stream = task_->GetEvent()->mutable_stream();
@@ -30,7 +30,7 @@
}
}
-void CaptureStreamInfo::AddOutput(const FloatAudioFrame& src) {
+void CaptureStreamInfo::AddOutput(const AudioFrameView<const float>& src) {
RTC_DCHECK(task_);
auto* stream = task_->GetEvent()->mutable_stream();
diff --git a/modules/audio_processing/aec_dump/capture_stream_info.h b/modules/audio_processing/aec_dump/capture_stream_info.h
index 9999c3f..91bb1fa 100644
--- a/modules/audio_processing/aec_dump/capture_stream_info.h
+++ b/modules/audio_processing/aec_dump/capture_stream_info.h
@@ -37,8 +37,8 @@
public:
explicit CaptureStreamInfo(std::unique_ptr<WriteToFileTask> task);
~CaptureStreamInfo();
- void AddInput(const FloatAudioFrame& src);
- void AddOutput(const FloatAudioFrame& src);
+ void AddInput(const AudioFrameView<const float>& src);
+ void AddOutput(const AudioFrameView<const float>& src);
void AddInput(const AudioFrame& frame);
void AddOutput(const AudioFrame& frame);
diff --git a/modules/audio_processing/aec_dump/mock_aec_dump.h b/modules/audio_processing/aec_dump/mock_aec_dump.h
index 6df6f28..8cfabdd 100644
--- a/modules/audio_processing/aec_dump/mock_aec_dump.h
+++ b/modules/audio_processing/aec_dump/mock_aec_dump.h
@@ -29,8 +29,10 @@
MOCK_METHOD1(WriteInitMessage,
void(const InternalAPMStreamsConfig& streams_config));
- MOCK_METHOD1(AddCaptureStreamInput, void(const FloatAudioFrame& src));
- MOCK_METHOD1(AddCaptureStreamOutput, void(const FloatAudioFrame& src));
+ MOCK_METHOD1(AddCaptureStreamInput,
+ void(const AudioFrameView<const float>& src));
+ MOCK_METHOD1(AddCaptureStreamOutput,
+ void(const AudioFrameView<const float>& src));
MOCK_METHOD1(AddCaptureStreamInput, void(const AudioFrame& frame));
MOCK_METHOD1(AddCaptureStreamOutput, void(const AudioFrame& frame));
MOCK_METHOD1(AddAudioProcessingState,
@@ -38,7 +40,8 @@
MOCK_METHOD0(WriteCaptureStreamMessage, void());
MOCK_METHOD1(WriteRenderStreamMessage, void(const AudioFrame& frame));
- MOCK_METHOD1(WriteRenderStreamMessage, void(const FloatAudioFrame& src));
+ MOCK_METHOD1(WriteRenderStreamMessage,
+ void(const AudioFrameView<const float>& src));
MOCK_METHOD1(WriteConfig, void(const InternalAPMConfig& config));
};
diff --git a/modules/audio_processing/agc2/BUILD.gn b/modules/audio_processing/agc2/BUILD.gn
new file mode 100644
index 0000000..8702e0f
--- /dev/null
+++ b/modules/audio_processing/agc2/BUILD.gn
@@ -0,0 +1,46 @@
+# Copyright (c) 2017 The WebRTC project authors. All Rights Reserved.
+#
+# 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.
+
+import("../../../webrtc.gni")
+
+rtc_source_set("agc2") {
+ sources = [
+ "agc2_common.h",
+ "fixed_gain_controller.cc",
+ "fixed_gain_controller.h",
+ ]
+
+ configs += [ "..:apm_debug_dump" ]
+
+ deps = [
+ "..:apm_logging",
+ "..:audio_frame_view",
+ "../../../api:array_view",
+ "../../../common_audio",
+ "../../../rtc_base:checks",
+ "../../../rtc_base:rtc_base_approved",
+ ]
+}
+
+rtc_source_set("fixed_digital_unittests") {
+ testonly = true
+ configs += [ "..:apm_debug_dump" ]
+
+ sources = [
+ "fixed_gain_controller_unittest.cc",
+ "vector_float_frame.cc",
+ "vector_float_frame.h",
+ ]
+ deps = [
+ ":agc2",
+ "..:apm_logging",
+ "..:audio_frame_view",
+ "../../../api:array_view",
+ "../../../rtc_base:rtc_base_tests_utils",
+ ]
+}
diff --git a/modules/audio_processing/agc2/agc2_common.h b/modules/audio_processing/agc2/agc2_common.h
new file mode 100644
index 0000000..c668d0a
--- /dev/null
+++ b/modules/audio_processing/agc2/agc2_common.h
@@ -0,0 +1,23 @@
+/*
+ * Copyright (c) 2018 The WebRTC project authors. All Rights Reserved.
+ *
+ * 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 MODULES_AUDIO_PROCESSING_AGC2_AGC2_COMMON_H_
+#define MODULES_AUDIO_PROCESSING_AGC2_AGC2_COMMON_H_
+
+namespace webrtc {
+
+constexpr float kMinSampleValue = -32768.f;
+constexpr float kMaxSampleValue = 32767.f;
+
+// TODO(aleloi): add the other constants as more AGC2 components are
+// added.
+} // namespace webrtc
+
+#endif // MODULES_AUDIO_PROCESSING_AGC2_AGC2_COMMON_H_
diff --git a/modules/audio_processing/agc2/fixed_gain_controller.cc b/modules/audio_processing/agc2/fixed_gain_controller.cc
new file mode 100644
index 0000000..a332365
--- /dev/null
+++ b/modules/audio_processing/agc2/fixed_gain_controller.cc
@@ -0,0 +1,90 @@
+/*
+ * Copyright (c) 2018 The WebRTC project authors. All Rights Reserved.
+ *
+ * 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.
+ */
+
+#include "modules/audio_processing/agc2/fixed_gain_controller.h"
+
+#include <algorithm>
+#include <cmath>
+
+#include "api/array_view.h"
+#include "common_audio/include/audio_util.h"
+#include "modules/audio_processing/agc2/agc2_common.h"
+#include "modules/audio_processing/logging/apm_data_dumper.h"
+#include "rtc_base/checks.h"
+#include "rtc_base/logging.h"
+#include "rtc_base/numerics/safe_minmax.h"
+
+namespace webrtc {
+namespace {
+
+// Returns true when the gain factor is so close to 1 that it would
+// not affect int16 samples.
+bool CloseToOne(float gain_factor) {
+ return 1.f - 1.f / kMaxSampleValue <= gain_factor &&
+ gain_factor <= 1.f + 1.f / kMaxSampleValue;
+}
+} // namespace
+
+FixedGainController::FixedGainController(ApmDataDumper* apm_data_dumper)
+ : apm_data_dumper_(apm_data_dumper) {
+ RTC_DCHECK_LT(0.f, gain_to_apply_);
+ RTC_DLOG(LS_INFO) << "Gain to apply: " << gain_to_apply_;
+}
+
+void FixedGainController::SetGain(float gain_to_apply_db) {
+ // Changes in gain_to_apply_ cause discontinuities. We assume
+ // gain_to_apply_ is set in the beginning of the call. If it is
+ // frequently changed, we should add interpolation between the
+ // values.
+ gain_to_apply_ = DbToRatio(gain_to_apply_db);
+}
+
+void FixedGainController::SetSampleRate(size_t sample_rate_hz) {
+ // TODO(aleloi): propagate the new sample rate to the GainCurveApplier.
+}
+
+void FixedGainController::EnableLimiter(bool enable_limiter) {
+ enable_limiter_ = enable_limiter;
+}
+
+void FixedGainController::Process(AudioFrameView<float> signal) {
+ // Apply fixed digital gain; interpolate if necessary. One of the
+ // planned usages of the FGC is to only use the limiter. In that
+ // case, the gain would be 1.0. Not doing the multiplications speeds
+ // it up considerably. Hence the check.
+ if (!CloseToOne(gain_to_apply_)) {
+ for (size_t k = 0; k < signal.num_channels(); ++k) {
+ rtc::ArrayView<float> channel_view = signal.channel(k);
+ for (auto& sample : channel_view) {
+ sample *= gain_to_apply_;
+ }
+ }
+ }
+
+ // Use the limiter (if configured to).
+ if (enable_limiter_) {
+ // TODO(aleloi): Process the signal with the
+ // GainCurveApplier. This will be done in the upcoming CLs.
+
+ // Dump data for debug.
+ const auto channel_view = signal.channel(0);
+ apm_data_dumper_->DumpRaw("agc2_fixed_digital_gain_curve_applier",
+ channel_view.size(), channel_view.data());
+ }
+
+ // Hard-clipping.
+ for (size_t k = 0; k < signal.num_channels(); ++k) {
+ rtc::ArrayView<float> channel_view = signal.channel(k);
+ for (auto& sample : channel_view) {
+ sample = rtc::SafeClamp(sample, kMinSampleValue, kMaxSampleValue);
+ }
+ }
+}
+} // namespace webrtc
diff --git a/modules/audio_processing/agc2/fixed_gain_controller.h b/modules/audio_processing/agc2/fixed_gain_controller.h
new file mode 100644
index 0000000..decc22e
--- /dev/null
+++ b/modules/audio_processing/agc2/fixed_gain_controller.h
@@ -0,0 +1,37 @@
+/*
+ * Copyright (c) 2018 The WebRTC project authors. All Rights Reserved.
+ *
+ * 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 MODULES_AUDIO_PROCESSING_AGC2_FIXED_GAIN_CONTROLLER_H_
+#define MODULES_AUDIO_PROCESSING_AGC2_FIXED_GAIN_CONTROLLER_H_
+
+#include "modules/audio_processing/include/audio_frame_view.h"
+
+namespace webrtc {
+class ApmDataDumper;
+
+class FixedGainController {
+ public:
+ explicit FixedGainController(ApmDataDumper* apm_data_dumper);
+
+ void Process(AudioFrameView<float> signal);
+
+ void SetGain(float gain_to_apply_db);
+ void SetSampleRate(size_t sample_rate_hz);
+ void EnableLimiter(bool enable_limiter);
+
+ private:
+ float gain_to_apply_ = 1.f;
+ ApmDataDumper* apm_data_dumper_;
+ bool enable_limiter_ = true;
+};
+
+} // namespace webrtc
+
+#endif // MODULES_AUDIO_PROCESSING_AGC2_FIXED_GAIN_CONTROLLER_H_
diff --git a/modules/audio_processing/agc2/fixed_gain_controller_unittest.cc b/modules/audio_processing/agc2/fixed_gain_controller_unittest.cc
new file mode 100644
index 0000000..9336fd7
--- /dev/null
+++ b/modules/audio_processing/agc2/fixed_gain_controller_unittest.cc
@@ -0,0 +1,109 @@
+/*
+ * Copyright (c) 2018 The WebRTC project authors. All Rights Reserved.
+ *
+ * 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.
+ */
+
+#include "modules/audio_processing/agc2/fixed_gain_controller.h"
+
+#include "api/array_view.h"
+#include "modules/audio_processing/agc2/vector_float_frame.h"
+#include "modules/audio_processing/logging/apm_data_dumper.h"
+#include "rtc_base/gunit.h"
+
+namespace webrtc {
+namespace {
+
+constexpr float kInputLevelLinear = 15000.f;
+
+constexpr float kGainToApplyDb = 15.f;
+
+float RunFixedGainControllerWithConstantInput(FixedGainController* fixed_gc,
+ const float input_level,
+ const size_t num_frames,
+ const int sample_rate) {
+ // Give time to the level etimator to converge.
+ for (size_t i = 0; i < num_frames; ++i) {
+ VectorFloatFrame vectors_with_float_frame(
+ 1, rtc::CheckedDivExact(sample_rate, 100), input_level);
+ fixed_gc->Process(vectors_with_float_frame.float_frame_view());
+ }
+
+ // Process the last frame with constant input level.
+ VectorFloatFrame vectors_with_float_frame_last(
+ 1, rtc::CheckedDivExact(sample_rate, 100), input_level);
+ fixed_gc->Process(vectors_with_float_frame_last.float_frame_view());
+
+ // Return the last sample from the last processed frame.
+ const auto channel =
+ vectors_with_float_frame_last.float_frame_view().channel(0);
+ return channel[channel.size() - 1];
+}
+ApmDataDumper test_data_dumper(0);
+
+FixedGainController CreateFixedGainController(float gain_to_apply,
+ size_t rate,
+ bool enable_limiter) {
+ FixedGainController fgc(&test_data_dumper);
+ fgc.SetGain(gain_to_apply);
+ fgc.SetSampleRate(gain_to_apply);
+ fgc.EnableLimiter(enable_limiter);
+ return fgc;
+}
+
+} // namespace
+
+TEST(AutomaticGainController2FixedDigital, CreateUseWithoutLimiter) {
+ const int kSampleRate = 48000;
+ FixedGainController fixed_gc =
+ CreateFixedGainController(kGainToApplyDb, kSampleRate, false);
+ VectorFloatFrame vectors_with_float_frame(
+ 1, rtc::CheckedDivExact(kSampleRate, 100), kInputLevelLinear);
+ auto float_frame = vectors_with_float_frame.float_frame_view();
+ fixed_gc.Process(float_frame);
+ const auto channel = float_frame.channel(0);
+ EXPECT_LT(kInputLevelLinear, channel[0]);
+}
+
+TEST(AutomaticGainController2FixedDigital, CreateUseWithLimiter) {
+ const int kSampleRate = 44000;
+ FixedGainController fixed_gc =
+ CreateFixedGainController(kGainToApplyDb, kSampleRate, true);
+ VectorFloatFrame vectors_with_float_frame(
+ 1, rtc::CheckedDivExact(kSampleRate, 100), kInputLevelLinear);
+ auto float_frame = vectors_with_float_frame.float_frame_view();
+ fixed_gc.Process(float_frame);
+ const auto channel = float_frame.channel(0);
+ EXPECT_LT(kInputLevelLinear, channel[0]);
+}
+
+TEST(AutomaticGainController2FixedDigital, GainShouldChangeOnSetGain) {
+ constexpr float input_level = 1000.f;
+ constexpr size_t num_frames = 5;
+ constexpr size_t kSampleRate = 8000;
+ constexpr float gain_db_no_change = 0.f;
+ constexpr float gain_db_factor_10 = 20.f;
+
+ FixedGainController fixed_gc_no_saturation =
+ CreateFixedGainController(gain_db_no_change, kSampleRate, false);
+
+ // Signal level is unchanged with 0 db gain.
+ EXPECT_FLOAT_EQ(
+ RunFixedGainControllerWithConstantInput(
+ &fixed_gc_no_saturation, input_level, num_frames, kSampleRate),
+ input_level);
+
+ fixed_gc_no_saturation.SetGain(gain_db_factor_10);
+
+ // +20db should increase signal by a factor of 10.
+ EXPECT_FLOAT_EQ(
+ RunFixedGainControllerWithConstantInput(
+ &fixed_gc_no_saturation, input_level, num_frames, kSampleRate),
+ input_level * 10);
+}
+
+} // namespace webrtc
diff --git a/modules/audio_processing/agc2/vector_float_frame.cc b/modules/audio_processing/agc2/vector_float_frame.cc
new file mode 100644
index 0000000..a70d815
--- /dev/null
+++ b/modules/audio_processing/agc2/vector_float_frame.cc
@@ -0,0 +1,39 @@
+/*
+ * Copyright (c) 2018 The WebRTC project authors. All Rights Reserved.
+ *
+ * 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.
+ */
+
+#include "modules/audio_processing/agc2/vector_float_frame.h"
+
+namespace webrtc {
+
+namespace {
+
+std::vector<float*> ConstructChannelPointers(
+ std::vector<std::vector<float>>* x) {
+ std::vector<float*> channel_ptrs;
+ for (auto& v : *x) {
+ channel_ptrs.push_back(v.data());
+ }
+ return channel_ptrs;
+}
+} // namespace
+
+VectorFloatFrame::VectorFloatFrame(int num_channels,
+ int samples_per_channel,
+ float start_value)
+ : channels_(num_channels,
+ std::vector<float>(samples_per_channel, start_value)),
+ channel_ptrs_(ConstructChannelPointers(&channels_)),
+ float_frame_view_(channel_ptrs_.data(),
+ channels_.size(),
+ samples_per_channel) {}
+
+VectorFloatFrame::~VectorFloatFrame() = default;
+
+} // namespace webrtc
diff --git a/modules/audio_processing/agc2/vector_float_frame.h b/modules/audio_processing/agc2/vector_float_frame.h
new file mode 100644
index 0000000..0e86089
--- /dev/null
+++ b/modules/audio_processing/agc2/vector_float_frame.h
@@ -0,0 +1,39 @@
+/*
+ * Copyright (c) 2018 The WebRTC project authors. All Rights Reserved.
+ *
+ * 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 MODULES_AUDIO_PROCESSING_AGC2_VECTOR_FLOAT_FRAME_H_
+#define MODULES_AUDIO_PROCESSING_AGC2_VECTOR_FLOAT_FRAME_H_
+
+#include <vector>
+
+#include "modules/audio_processing/include/audio_frame_view.h"
+
+namespace webrtc {
+
+// A construct consisting of a multi-channel audio frame, and a FloatFrame view
+// of it.
+class VectorFloatFrame {
+ public:
+ VectorFloatFrame(int num_channels,
+ int samples_per_channel,
+ float start_value);
+ const AudioFrameView<float>& float_frame_view() { return float_frame_view_; }
+
+ ~VectorFloatFrame();
+
+ private:
+ std::vector<std::vector<float>> channels_;
+ std::vector<float*> channel_ptrs_;
+ AudioFrameView<float> float_frame_view_;
+};
+
+} // namespace webrtc
+
+#endif // MODULES_AUDIO_PROCESSING_AGC2_VECTOR_FLOAT_FRAME_H_
diff --git a/modules/audio_processing/audio_processing_impl.cc b/modules/audio_processing/audio_processing_impl.cc
index 6275726..a9e6b05 100644
--- a/modules/audio_processing/audio_processing_impl.cc
+++ b/modules/audio_processing/audio_processing_impl.cc
@@ -20,7 +20,6 @@
#include "common_audio/signal_processing/include/signal_processing_library.h"
#include "modules/audio_processing/aec/aec_core.h"
#include "modules/audio_processing/agc/agc_manager_direct.h"
-#include "modules/audio_processing/agc2/gain_controller2.h"
#include "modules/audio_processing/audio_buffer.h"
#include "modules/audio_processing/beamformer/nonlinear_beamformer.h"
#include "modules/audio_processing/common.h"
@@ -28,6 +27,7 @@
#include "modules/audio_processing/echo_control_mobile_impl.h"
#include "modules/audio_processing/gain_control_for_experimental_agc.h"
#include "modules/audio_processing/gain_control_impl.h"
+#include "modules/audio_processing/gain_controller2.h"
#include "modules/audio_processing/logging/apm_data_dumper.h"
#include "rtc_base/checks.h"
#include "rtc_base/logging.h"
@@ -1425,7 +1425,7 @@
const size_t num_channels =
formats_.api_format.reverse_input_stream().num_channels();
aec_dump_->WriteRenderStreamMessage(
- FloatAudioFrame(src, num_channels, channel_size));
+ AudioFrameView<const float>(src, num_channels, channel_size));
}
render_.render_audio->CopyFrom(src,
formats_.api_format.reverse_input_stream());
@@ -1993,7 +1993,7 @@
const size_t channel_size = formats_.api_format.input_stream().num_frames();
const size_t num_channels = formats_.api_format.input_stream().num_channels();
aec_dump_->AddCaptureStreamInput(
- FloatAudioFrame(src, num_channels, channel_size));
+ AudioFrameView<const float>(src, num_channels, channel_size));
RecordAudioProcessingState();
}
@@ -2013,8 +2013,8 @@
const size_t channel_size = formats_.api_format.output_stream().num_frames();
const size_t num_channels =
formats_.api_format.output_stream().num_channels();
- aec_dump_->AddCaptureStreamOutput(
- FloatAudioFrame(processed_capture_stream, num_channels, channel_size));
+ aec_dump_->AddCaptureStreamOutput(AudioFrameView<const float>(
+ processed_capture_stream, num_channels, channel_size));
aec_dump_->WriteCaptureStreamMessage();
}
diff --git a/modules/audio_processing/agc2/gain_controller2.cc b/modules/audio_processing/gain_controller2.cc
similarity index 69%
rename from modules/audio_processing/agc2/gain_controller2.cc
rename to modules/audio_processing/gain_controller2.cc
index b02aa59..aa866c1 100644
--- a/modules/audio_processing/agc2/gain_controller2.cc
+++ b/modules/audio_processing/gain_controller2.cc
@@ -8,16 +8,13 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-#include "modules/audio_processing/agc2/gain_controller2.h"
+#include "modules/audio_processing/gain_controller2.h"
-#include <cmath>
-
-#include "common_audio/include/audio_util.h"
#include "modules/audio_processing/audio_buffer.h"
+#include "modules/audio_processing/include/audio_frame_view.h"
#include "modules/audio_processing/logging/apm_data_dumper.h"
#include "rtc_base/atomicops.h"
#include "rtc_base/checks.h"
-#include "rtc_base/numerics/safe_minmax.h"
namespace webrtc {
@@ -26,8 +23,7 @@
GainController2::GainController2()
: data_dumper_(
new ApmDataDumper(rtc::AtomicOps::Increment(&instance_count_))),
- sample_rate_hz_(AudioProcessing::kSampleRate48kHz),
- fixed_gain_(1.f) {}
+ gain_controller_(data_dumper_.get()) {}
GainController2::~GainController2() = default;
@@ -36,28 +32,23 @@
sample_rate_hz == AudioProcessing::kSampleRate16kHz ||
sample_rate_hz == AudioProcessing::kSampleRate32kHz ||
sample_rate_hz == AudioProcessing::kSampleRate48kHz);
- sample_rate_hz_ = sample_rate_hz;
+ gain_controller_.SetSampleRate(sample_rate_hz);
data_dumper_->InitiateNewSetOfRecordings();
- data_dumper_->DumpRaw("sample_rate_hz", sample_rate_hz_);
- data_dumper_->DumpRaw("fixed_gain_linear", fixed_gain_);
+ data_dumper_->DumpRaw("sample_rate_hz", sample_rate_hz);
}
void GainController2::Process(AudioBuffer* audio) {
- if (fixed_gain_ == 1.f)
- return;
-
- for (size_t k = 0; k < audio->num_channels(); ++k) {
- for (size_t j = 0; j < audio->num_frames(); ++j) {
- audio->channels_f()[k][j] = rtc::SafeClamp(
- fixed_gain_ * audio->channels_f()[k][j], -32768.f, 32767.f);
- }
- }
+ AudioFrameView<float> float_frame(audio->channels_f(), audio->num_channels(),
+ audio->num_frames());
+ gain_controller_.Process(float_frame);
}
void GainController2::ApplyConfig(
const AudioProcessing::Config::GainController2& config) {
RTC_DCHECK(Validate(config));
- fixed_gain_ = DbToRatio(config.fixed_gain_db);
+ config_ = config;
+ gain_controller_.SetGain(config_.fixed_gain_db);
+ gain_controller_.EnableLimiter(config_.enable_limiter);
}
bool GainController2::Validate(
@@ -69,7 +60,8 @@
const AudioProcessing::Config::GainController2& config) {
std::stringstream ss;
ss << "{enabled: " << (config.enabled ? "true" : "false") << ", "
- << "fixed_gain_dB: " << config.fixed_gain_db << "}";
+ << "fixed_gain_dB: " << config.fixed_gain_db << ", "
+ << "enable_limiter: " << (config.enable_limiter ? "true" : "false") << "}";
return ss.str();
}
diff --git a/modules/audio_processing/agc2/gain_controller2.h b/modules/audio_processing/gain_controller2.h
similarity index 80%
rename from modules/audio_processing/agc2/gain_controller2.h
rename to modules/audio_processing/gain_controller2.h
index 1170687..6de4564 100644
--- a/modules/audio_processing/agc2/gain_controller2.h
+++ b/modules/audio_processing/gain_controller2.h
@@ -8,12 +8,13 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-#ifndef MODULES_AUDIO_PROCESSING_AGC2_GAIN_CONTROLLER2_H_
-#define MODULES_AUDIO_PROCESSING_AGC2_GAIN_CONTROLLER2_H_
+#ifndef MODULES_AUDIO_PROCESSING_GAIN_CONTROLLER2_H_
+#define MODULES_AUDIO_PROCESSING_GAIN_CONTROLLER2_H_
#include <memory>
#include <string>
+#include "modules/audio_processing/agc2/fixed_gain_controller.h"
#include "modules/audio_processing/include/audio_processing.h"
#include "rtc_base/constructormagic.h"
@@ -24,8 +25,6 @@
// Gain Controller 2 aims to automatically adjust levels by acting on the
// microphone gain and/or applying digital gain.
-//
-// Temporarily implements a fixed gain mode with hard-clipping.
class GainController2 {
public:
GainController2();
@@ -42,12 +41,12 @@
private:
static int instance_count_;
std::unique_ptr<ApmDataDumper> data_dumper_;
- int sample_rate_hz_;
- float fixed_gain_;
+ FixedGainController gain_controller_;
+ AudioProcessing::Config::GainController2 config_;
RTC_DISALLOW_COPY_AND_ASSIGN(GainController2);
};
} // namespace webrtc
-#endif // MODULES_AUDIO_PROCESSING_AGC2_GAIN_CONTROLLER2_H_
+#endif // MODULES_AUDIO_PROCESSING_GAIN_CONTROLLER2_H_
diff --git a/modules/audio_processing/agc2/gain_controller2_unittest.cc b/modules/audio_processing/gain_controller2_unittest.cc
similarity index 91%
rename from modules/audio_processing/agc2/gain_controller2_unittest.cc
rename to modules/audio_processing/gain_controller2_unittest.cc
index 46f654d..0c3b383 100644
--- a/modules/audio_processing/agc2/gain_controller2_unittest.cc
+++ b/modules/audio_processing/gain_controller2_unittest.cc
@@ -11,8 +11,8 @@
#include <algorithm>
#include "api/array_view.h"
-#include "modules/audio_processing/agc2/gain_controller2.h"
#include "modules/audio_processing/audio_buffer.h"
+#include "modules/audio_processing/gain_controller2.h"
#include "rtc_base/checks.h"
#include "test/gtest.h"
@@ -61,11 +61,12 @@
config.fixed_gain_db = 5.f;
config.enabled = false;
- EXPECT_EQ("{enabled: false, fixed_gain_dB: 5}",
+ config.enable_limiter = true;
+ EXPECT_EQ("{enabled: false, fixed_gain_dB: 5, enable_limiter: true}",
GainController2::ToString(config));
config.enabled = true;
- EXPECT_EQ("{enabled: true, fixed_gain_dB: 5}",
+ EXPECT_EQ("{enabled: true, fixed_gain_dB: 5, enable_limiter: true}",
GainController2::ToString(config));
}
@@ -81,7 +82,7 @@
AudioProcessing::Config::GainController2 config;
// Check that samples are not modified when the fixed gain is 0 dB.
- ASSERT_EQ(config.fixed_gain_db, 0.f);
+ config.fixed_gain_db = 0.f;
gain_controller2->ApplyConfig(config);
gain_controller2->Process(&ab);
EXPECT_EQ(ab.channels_f()[0][0], sample_value);
diff --git a/modules/audio_processing/include/aec_dump.h b/modules/audio_processing/include/aec_dump.h
index 0c8d227..2035bf4 100644
--- a/modules/audio_processing/include/aec_dump.h
+++ b/modules/audio_processing/include/aec_dump.h
@@ -16,6 +16,7 @@
#include <vector>
#include "api/array_view.h"
+#include "modules/audio_processing/include/audio_frame_view.h"
namespace webrtc {
@@ -65,37 +66,6 @@
size_t render_output_num_channels = 0;
};
-// Class to pass audio data in float** format. This is to avoid
-// dependence on AudioBuffer, and avoid problems associated with
-// rtc::ArrayView<rtc::ArrayView>.
-class FloatAudioFrame {
- public:
- // |num_channels| and |channel_size| describe the float**
- // |audio_samples|. |audio_samples| is assumed to point to a
- // two-dimensional |num_channels * channel_size| array of floats.
- FloatAudioFrame(const float* const* audio_samples,
- size_t num_channels,
- size_t channel_size)
- : audio_samples_(audio_samples),
- num_channels_(num_channels),
- channel_size_(channel_size) {}
-
- FloatAudioFrame() = delete;
-
- size_t num_channels() const { return num_channels_; }
-
- rtc::ArrayView<const float> channel(size_t idx) const {
- RTC_DCHECK_LE(0, idx);
- RTC_DCHECK_LE(idx, num_channels_);
- return rtc::ArrayView<const float>(audio_samples_[idx], channel_size_);
- }
-
- private:
- const float* const* audio_samples_;
- size_t num_channels_;
- size_t channel_size_;
-};
-
// An interface for recording configuration and input/output streams
// of the Audio Processing Module. The recordings are called
// 'aec-dumps' and are stored in a protobuf format defined in
@@ -122,8 +92,10 @@
// Logs Event::Type STREAM message. To log an input/output pair,
// call the AddCapture* and AddAudioProcessingState methods followed
// by a WriteCaptureStreamMessage call.
- virtual void AddCaptureStreamInput(const FloatAudioFrame& src) = 0;
- virtual void AddCaptureStreamOutput(const FloatAudioFrame& src) = 0;
+ virtual void AddCaptureStreamInput(
+ const AudioFrameView<const float>& src) = 0;
+ virtual void AddCaptureStreamOutput(
+ const AudioFrameView<const float>& src) = 0;
virtual void AddCaptureStreamInput(const AudioFrame& frame) = 0;
virtual void AddCaptureStreamOutput(const AudioFrame& frame) = 0;
virtual void AddAudioProcessingState(const AudioProcessingState& state) = 0;
@@ -131,7 +103,8 @@
// Logs Event::Type REVERSE_STREAM message.
virtual void WriteRenderStreamMessage(const AudioFrame& frame) = 0;
- virtual void WriteRenderStreamMessage(const FloatAudioFrame& src) = 0;
+ virtual void WriteRenderStreamMessage(
+ const AudioFrameView<const float>& src) = 0;
// Logs Event::Type CONFIG message.
virtual void WriteConfig(const InternalAPMConfig& config) = 0;
diff --git a/modules/audio_processing/include/audio_frame_view.h b/modules/audio_processing/include/audio_frame_view.h
new file mode 100644
index 0000000..86e593a
--- /dev/null
+++ b/modules/audio_processing/include/audio_frame_view.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2018 The WebRTC project authors. All Rights Reserved.
+ *
+ * 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 MODULES_AUDIO_PROCESSING_INCLUDE_AUDIO_FRAME_VIEW_H_
+#define MODULES_AUDIO_PROCESSING_INCLUDE_AUDIO_FRAME_VIEW_H_
+
+#include "api/array_view.h"
+
+// Class to pass audio data in T** format, where T is a numeric type.
+template <class T>
+class AudioFrameView {
+ public:
+ // |num_channels| and |channel_size| describe the T**
+ // |audio_samples|. |audio_samples| is assumed to point to a
+ // two-dimensional |num_channels * channel_size| array of floats.
+ AudioFrameView(T* const* audio_samples,
+ size_t num_channels,
+ size_t channel_size)
+ : audio_samples_(audio_samples),
+ num_channels_(num_channels),
+ channel_size_(channel_size) {}
+
+ AudioFrameView() = delete;
+
+ size_t num_channels() const { return num_channels_; }
+
+ size_t samples_per_channel() const { return channel_size_; }
+
+ rtc::ArrayView<T> channel(size_t idx) {
+ RTC_DCHECK_LE(0, idx);
+ RTC_DCHECK_LE(idx, num_channels_);
+ return rtc::ArrayView<T>(audio_samples_[idx], channel_size_);
+ }
+
+ rtc::ArrayView<const T> channel(size_t idx) const {
+ RTC_DCHECK_LE(0, idx);
+ RTC_DCHECK_LE(idx, num_channels_);
+ return rtc::ArrayView<const T>(audio_samples_[idx], channel_size_);
+ }
+
+ private:
+ T* const* audio_samples_;
+ size_t num_channels_;
+ size_t channel_size_;
+};
+
+#endif // MODULES_AUDIO_PROCESSING_INCLUDE_AUDIO_FRAME_VIEW_H_
diff --git a/modules/audio_processing/include/audio_processing.h b/modules/audio_processing/include/audio_processing.h
index 50ec430..06c1f4a 100644
--- a/modules/audio_processing/include/audio_processing.h
+++ b/modules/audio_processing/include/audio_processing.h
@@ -284,6 +284,7 @@
struct GainController2 {
bool enabled = false;
float fixed_gain_db = 0.f;
+ bool enable_limiter = true;
} gain_controller2;
// Explicit copy assignment implementation to avoid issues with memory
