Conversational speech tool, simualtor + unit tests

webrtc/modules/audio_processing/test/conversational_speech/simulator.cc

Index: webrtc/modules/audio_processing/test/conversational_speech/simulator.cc
diff --git a/webrtc/modules/audio_processing/test/conversational_speech/simulator.cc b/webrtc/modules/audio_processing/test/conversational_speech/simulator.cc
new file mode 100644
index 0000000000000000000000000000000000000000..0225a0475d6c15070de9dd29e92cb9bb74da52d6
--- /dev/null
+++ b/webrtc/modules/audio_processing/test/conversational_speech/simulator.cc
@@ -0,0 +1,219 @@
+/*
+ *  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.
+ */
+
+#include "webrtc/modules/audio_processing/test/conversational_speech/simulator.h"
+
+#include <set>
+#include <utility>
+#include <vector>
+
+#include "webrtc/base/array_view.h"
+#include "webrtc/base/constructormagic.h"
+#include "webrtc/base/logging.h"
+#include "webrtc/base/pathutils.h"
+#include "webrtc/base/ptr_util.h"
+#include "webrtc/common_audio/wav_file.h"
+#include "webrtc/modules/audio_processing/test/conversational_speech/wavreader_interface.h"
+
+namespace webrtc {
+namespace test {
+namespace {
+
+using conversational_speech::MultiEndCall;
+using conversational_speech::SpeakerOutputFilePaths;
+using conversational_speech::WavReaderInterface;
+
+// Combines output path and speaker names to define the output file paths for
+// the near-end and far=end audio tracks.
+std::unique_ptr<std::map<std::string, SpeakerOutputFilePaths>>
+    InitSpeakerOutputFilePaths(const std::set<std::string>& speaker_names,
+                               const std::string& output_path) {
+  // Create map.
+  auto speaker_output_file_paths_map = rtc::MakeUnique<
+      std::map<std::string, SpeakerOutputFilePaths>>();
+
+  // Add near-end and far-end output paths into the map.
+  for (const auto& speaker_name : speaker_names) {
+    const rtc::Pathname near_end_path(
+        output_path, "s_" + speaker_name + "-near_end.wav");
+    LOG(LS_VERBOSE) << "creating " << near_end_path.pathname();

[minyue-webrtc, 2017/05/16 15:05:17]

I'd like full sentence in the log

[AleBzk, 2017/05/17 12:49:38]

Done.

+
+    const rtc::Pathname far_end_path(
+        output_path, "s_" + speaker_name + "-far_end.wav");
+    LOG(LS_VERBOSE) << "creating " << far_end_path.pathname();
+
+    // Add to map.
+    speaker_output_file_paths_map->emplace(
+        std::piecewise_construct,

[minyue-webrtc, 2017/05/16 15:05:16]

I am not a fan of this, ok, I think I understand now.

[AleBzk, 2017/05/17 12:49:38]

Acknowledged.

+        std::forward_as_tuple(speaker_name),
+        std::forward_as_tuple(near_end_path.pathname(),
+                              far_end_path.pathname()));
+  }
+
+  return speaker_output_file_paths_map;
+}
+
+// Class that provides one WavWriter for the near-end and one for the far-end
+// output track of a speaker.
+class SpeakerWavWriters {
+ public:
+  SpeakerWavWriters(
+      const SpeakerOutputFilePaths& output_file_paths, int sample_rate)
+          : near_end_(output_file_paths.near_end, sample_rate, 1u),
+            far_end_(output_file_paths.far_end, sample_rate, 1u) {}
+  WavWriter& near_end() {

[minyue-webrtc, 2017/05/16 15:05:17]

WavWriter* near_end_wav_writer()

+    return near_end_;
+  }
+  WavWriter& far_end() {

[minyue-webrtc, 2017/05/16 15:05:17]

WavWriter* far_end_wav_writer()

[AleBzk, 2017/05/17 12:49:38]

Done.

+    return far_end_;
+  }
+ private:
+  WavWriter near_end_;

[minyue-webrtc, 2017/05/16 15:05:17]

near_end_wav_writer_

[AleBzk, 2017/05/17 12:49:38]

Done.

[minyue-webrtc, 2017/05/19 12:46:51]

I'd prefer change the var name to near_end_wav_writer_

[AleBzk, 2017/05/19 15:45:04]

Done.

+  WavWriter far_end_;

[minyue-webrtc, 2017/05/16 15:05:17]

far_end_wav_writer_

[AleBzk, 2017/05/17 12:49:38]

Done.

[minyue-webrtc, 2017/05/19 12:46:51]

same here

[AleBzk, 2017/05/19 15:45:04]

Done.

+};
+
+// Initializes one WavWriter instance for each speaker and both the near-end and
+// far-end output tracks.
+std::unique_ptr<std::map<std::string, SpeakerWavWriters>>
+    InitSpeakersWavWriters(const std::map<std::string, SpeakerOutputFilePaths>&
+                           speaker_output_file_paths, int sample_rate) {
+  // Create map.
+  auto speaker_wav_writers_map = rtc::MakeUnique<
+      std::map<std::string, SpeakerWavWriters>>();
+
+  // Add SpeakerWavWriters instance into the map.
+  for (auto it = speaker_output_file_paths.begin();
+      it != speaker_output_file_paths.end(); ++it) {
+    speaker_wav_writers_map->emplace(
+        std::piecewise_construct,
+        std::forward_as_tuple(it->first),
+        std::forward_as_tuple(it->second, sample_rate));
+  }
+
+  return speaker_wav_writers_map;
+}
+
+// Reads all the samples for each audio track.
+std::unique_ptr<std::map<std::string, std::vector<int16_t>>> PreloadAudioTracks(
+    const std::map<std::string, std::unique_ptr<WavReaderInterface>>&
+        audiotrack_readers) {
+  // Create map.
+  auto audiotracks_map = rtc::MakeUnique<
+      std::map<std::string, std::vector<int16_t>>>();
+
+  // Add audio track vectors.
+  for (auto it = audiotrack_readers.begin(); it != audiotrack_readers.end();
+      ++it) {
+    // Add map entry.
+    audiotracks_map->emplace(
+        std::piecewise_construct,
+        std::forward_as_tuple(it->first),
+        std::forward_as_tuple(it->second->NumSamples()));
+
+    // Read samples.
+    it->second->ReadInt16Samples(audiotracks_map->at(it->first));
+  }
+
+  return audiotracks_map;
+}
+
+// Writes all the values in |source_samples| via |wav_writer|. If the number of
+// previously written samples in |wav_writer| is less than |interval_begin|, it
+// adds zeros as left padding. The padding corresponds to intervals during which
+// a speaker is not active.
+void PadLeftWriteChunk(rtc::ArrayView<const int16_t> source_samples,
+                       size_t interval_begin, WavWriter* wav_writer) {
+  // Add left padding.
+  RTC_CHECK(wav_writer);
+  RTC_CHECK_GE(interval_begin, wav_writer->num_samples());
+  size_t padding_size = interval_begin - wav_writer->num_samples();
+  if (padding_size > 0) {

[minyue-webrtc, 2017/05/16 15:05:17]

I prefer "!=0": a bit easier to read

[AleBzk, 2017/05/17 12:49:38]

Done.

+    const std::vector<int16_t> padding(padding_size, 0);
+    wav_writer->WriteSamples(padding.data(), padding_size);
+  }
+
+  // Write source samples.
+  wav_writer->WriteSamples(source_samples.data(), source_samples.size());
+}
+
+// Appends zeros via |wav_writer|. The number of zeros is always non-negative
+// and equal to the difference between the previously written samples and
+// |pad_samples|.
+void PadRightWrite(WavWriter* wav_writer, size_t pad_samples) {
+  RTC_CHECK(wav_writer);
+  RTC_CHECK_GE(pad_samples, wav_writer->num_samples());
+  size_t padding_size = pad_samples - wav_writer->num_samples();
+  if (padding_size > 0) {

[minyue-webrtc, 2017/05/16 15:05:17]

!= 0

[AleBzk, 2017/05/17 12:49:38]

Done.

+    const std::vector<int16_t> padding(padding_size, 0);
+    wav_writer->WriteSamples(padding.data(), padding_size);
+  }
+}
+
+}  // namespace
+
+namespace conversational_speech {
+
+std::unique_ptr<std::map<std::string, SpeakerOutputFilePaths>> Simulate(
+    const MultiEndCall& multiend_call, const std::string& output_path) {
+  // Set output file paths and initialize wav writers.
+  const auto& speaker_names = multiend_call.speaker_names();
+  auto speaker_output_file_paths = InitSpeakerOutputFilePaths(
+      speaker_names, output_path);
+  auto speakers_wav_writers = InitSpeakersWavWriters(
+      *speaker_output_file_paths, multiend_call.sample_rate());
+
+  // Preload all the input audio tracks.
+  const auto& audiotrack_readers = multiend_call.audiotrack_readers();
+  auto audiotracks = PreloadAudioTracks(audiotrack_readers);
+
+  // TODO(alessiob): When speaker_names.size() == 2, near-end and far-end
+  // across the 2 speakers are symmetric; hence, the code below could be
+  // replaced by only creating the near-end or the far-end. However, this would
+  // require to split the unit tests and document the behavior in README.md.
+  // In practice, it should not be an issue since the files are not expected to
+  // be signinificant.
+
+  // Write near-end and far-end output tracks.
+  const auto& speaking_turns = multiend_call.speaking_turns();
+  for (const auto& speaking_turn : speaking_turns) {

[minyue-webrtc, 2017/05/16 15:05:16]

I think you may put multiend_call.speaking_turns() in the for()

[AleBzk, 2017/05/17 12:49:38]

Done.

+    const std::string& active_speaker_name = speaking_turn.speaker_name;
+    auto source_audiotrack = audiotracks->at(
+        speaking_turn.audiotrack_file_name);
+
+    // Write active speaker's chunk to active speaker's near-end.
+    PadLeftWriteChunk(source_audiotrack, speaking_turn.begin,
+                      &speakers_wav_writers->at(

[minyue-webrtc, 2017/05/16 15:05:17]

remove & if you return * with near_end() 

there are other similar places.

[AleBzk, 2017/05/17 12:49:38]

Done.

+                          active_speaker_name).near_end());
+
+    // Write active speaker's chunk to other participants' far-ends.
+    for (const std::string& speaker_name : speaker_names) {
+      if (speaker_name == active_speaker_name)
+        continue;
+      PadLeftWriteChunk(source_audiotrack, speaking_turn.begin,
+                        &speakers_wav_writers->at(speaker_name).far_end());
+    }
+  }
+
+  // Finalize all the output tracks with right padding.
+  // This is required to make all the output tracks duration equal.
+  size_t duration_samples = multiend_call.total_duration_samples();
+  for (const std::string& speaker_name : speaker_names) {
+    PadRightWrite(&speakers_wav_writers->at(speaker_name).near_end(),
+                  duration_samples);
+    PadRightWrite(&speakers_wav_writers->at(speaker_name).far_end(),
+                  duration_samples);
+  }
+
+  return speaker_output_file_paths;
+}
+
+}  // namespace conversational_speech
+}  // namespace test
+}  // namespace webrtc