Add IntelligibilityEnhancer support to audioproc_float

webrtc/modules/audio_processing/test/audioproc_float.cc

 /*
  *  Copyright (c) 2014 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.
  */
 
@@ skipping 24 matching lines @@
 }
 
 }  // namespace
 
 DEFINE_string(dump, "", "Name of the aecdump debug file to read from.");
 DEFINE_string(i, "", "Name of the capture input stream file to read from.");
 DEFINE_string(
     o,
     "out.wav",
     "Name of the output file to write the processed capture stream to.");
+DEFINE_string(ri, "", "Name of the render input stream file to read from.");
+DEFINE_string(
+    ro,
+    "out_reverse.wav",
+    "Name of the output file to write the processed render stream to.");
 DEFINE_int32(out_channels, 1, "Number of output channels.");
 const bool out_channels_dummy =
     google::RegisterFlagValidator(&FLAGS_out_channels, &ValidateOutChannels);
+DEFINE_int32(rev_out_channels, 1, "Number of reverse output channels.");
+const bool rev_out_channels_dummy =
+    google::RegisterFlagValidator(&FLAGS_rev_out_channels,
+                                  &ValidateOutChannels);
 DEFINE_int32(out_sample_rate, 48000, "Output sample rate in Hz.");
+DEFINE_int32(rev_out_sample_rate, 48000, "Reverse output sample rate in Hz.");
 DEFINE_string(mic_positions, "",
     "Space delimited cartesian coordinates of microphones in meters. "
     "The coordinates of each point are contiguous. "
     "For a two element array: \"x1 y1 z1 x2 y2 z2\"");
 DEFINE_double(
     target_angle_degrees,
     90,
     "The azimuth of the target in degrees. Only applies to beamforming.");
 
 DEFINE_bool(aec, false, "Enable echo cancellation.");
@@ skipping 11 matching lines @@
 
 namespace webrtc {
 namespace {
 
 const int kChunksPerSecond = 100;
 const char kUsage[] =
     "Command-line tool to run audio processing on WAV files. Accepts either\n"
     "an input capture WAV file or protobuf debug dump and writes to an output\n"
     "WAV file.\n"
     "\n"
-    "All components are disabled by default. If any bi-directional components\n"
-    "are enabled, only debug dump files are permitted.";
+    "All components are disabled by default. If the AEC is enabled, only\n"

[peah-webrtc, 2016/03/18 21:57:17]

Actually, I'd rather you fully remove this line about the AEC. It is only valid
due to lines 104-106, and the FLAGS_all setting anyway may turn on the AEC when
operating on wav-files. And apart from not being able to set the proper default
settings for the AEC, I see now reason for why the AEC cannot already be run
using wav-files.

[aluebs-webrtc, 2016/03/18 23:39:29]

Agreed. Done.

+    "debug dump files are permitted.";
 
 }  // namespace
 
 int main(int argc, char* argv[]) {
   google::SetUsageMessage(kUsage);
   google::ParseCommandLineFlags(&argc, &argv, true);
 
   if (!((FLAGS_i.empty()) ^ (FLAGS_dump.empty()))) {
     fprintf(stderr,
             "An input file must be specified with either -i or -dump.\n");
     return 1;
   }
-  if (FLAGS_dump.empty() && (FLAGS_aec || FLAGS_ie)) {
-    fprintf(stderr, "-aec and -ie require a -dump file.\n");
-    return 1;
-  }
-  if (FLAGS_ie) {
-    fprintf(stderr,
-            "FIXME(ajm): The intelligibility enhancer output is not dumped.\n");
+  if (FLAGS_dump.empty() && FLAGS_aec) {

[peah-webrtc, 2016/03/18 21:57:17]

Please remove this statement. The statement is not more true for the AEC than
for NS, AGC or any other component. The test program is not at all yet ready to
operate properly on any of those and and I think that it is then better not to
give the impression that it works on wav files for anything but the AEC.

[aluebs-webrtc, 2016/03/18 23:39:29]

Done. If by "not at all yet ready to operate properly" you mean that some
settings are missing, I agree. Or is there something else missing?

[peah-webrtc, 2016/03/19 01:00:16]

There is afaics lots of things missing:
AEC:
set_suppression_level
set_stream_drift_compensation
delay agnostic mode
extended filter mode

AECM
routing mode
comfort noise enabled
set echo path

GainControl
set_stream_analog_level
set_mode
set_target_level_dbfs
set_compression_gain_db
enable_limiter
set_analog_level_limits

I would say that basically all of the submodule-specific settings are missing
apart from setting the noise suppression level which is the only one that is
actually implemented. The settings listed above are the settings that are
publicly available in the APM submodule API and which therefore a test
application must support.

[aluebs-webrtc, 2016/03/19 02:09:55]

Thank you for doing that analysis. I agree we should support these settings. And
I don't think it is a lot of work to do so.

+    fprintf(stderr, "-aec requires a -dump file.\n");
     return 1;
   }
 
   test::TraceToStderr trace_to_stderr(true);
   Config config;
   if (FLAGS_bf || FLAGS_all) {
     if (FLAGS_mic_positions.empty()) {
       fprintf(stderr, "-mic_positions must be specified when -bf is used.\n");
       return 1;
     }
@@ skipping 17 matching lines @@
   }
   ap->set_stream_key_pressed(FLAGS_ts);
 
   std::unique_ptr<AudioFileProcessor> processor;
   auto out_file = std::unique_ptr<WavWriter>(new WavWriter(
       FLAGS_o, FLAGS_out_sample_rate, static_cast<size_t>(FLAGS_out_channels)));
   std::cout << FLAGS_o << ": " << out_file->FormatAsString() << std::endl;
   if (FLAGS_dump.empty()) {
     auto in_file = std::unique_ptr<WavReader>(new WavReader(FLAGS_i));
     std::cout << FLAGS_i << ": " << in_file->FormatAsString() << std::endl;
-    processor.reset(new WavFileProcessor(std::move(ap), std::move(in_file),
-                                         std::move(out_file)));
+    std::unique_ptr<WavReader> reverse_in_file;
+    std::unique_ptr<WavWriter> reverse_out_file;
+    if (!FLAGS_ri.empty()) {
+      reverse_in_file.reset(new WavReader(FLAGS_ri));
+      reverse_out_file.reset(new WavWriter(
+          FLAGS_ro,
+          FLAGS_rev_out_sample_rate,
+          static_cast<size_t>(FLAGS_rev_out_channels)));
+      std::cout << FLAGS_ri << ": "
+                << reverse_in_file->FormatAsString() << std::endl;
+      std::cout << FLAGS_ro << ": "
+                << reverse_out_file->FormatAsString() << std::endl;
+    }
+    processor.reset(new WavFileProcessor(std::move(ap),
+                                         std::move(in_file),
+                                         std::move(out_file),
+                                         std::move(reverse_in_file),
+                                         std::move(reverse_out_file)));
 
   } else {
     processor.reset(new AecDumpFileProcessor(
         std::move(ap), fopen(FLAGS_dump.c_str(), "rb"), std::move(out_file)));
   }
 
   int num_chunks = 0;
   while (processor->ProcessChunk()) {
     trace_to_stderr.SetTimeSeconds(num_chunks * 1.f / kChunksPerSecond);
     ++num_chunks;
@@ skipping 11 matching lines @@
   }
 
   return 0;
 }
 
 }  // namespace webrtc
 
 int main(int argc, char* argv[]) {
   return webrtc::main(argc, argv);
 }