chrome/browser/media/chrome_webrtc_audio_quality_browsertest.cc

   1 // Copyright 2013 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include <ctime>
   6
   7 #include "base/command_line.h"
   8 #include "base/file_util.h"
   9 #include "base/path_service.h"
  10 #include "base/process/launch.h"
  11 #include "base/scoped_native_library.h"
  12 #include "base/strings/stringprintf.h"
  13 #include "base/win/windows_version.h"
  14 #include "chrome/browser/media/webrtc_browsertest_base.h"
  15 #include "chrome/browser/media/webrtc_browsertest_common.h"
  16 #include "chrome/browser/profiles/profile.h"
  17 #include "chrome/browser/ui/browser.h"
  18 #include "chrome/browser/ui/browser_tabstrip.h"
  19 #include "chrome/browser/ui/tabs/tab_strip_model.h"
  20 #include "chrome/common/chrome_paths.h"
  21 #include "chrome/common/chrome_switches.h"
  22 #include "chrome/test/base/ui_test_utils.h"
  23 #include "content/public/test/browser_test_utils.h"
  24 #include "net/test/embedded_test_server/embedded_test_server.h"
  25 #include "testing/perf/perf_test.h"
  26
  27 // These are relative to the reference file dir defined by
  28 // webrtc_browsertest_common.h (i.e. chrome/test/data/webrtc/resources).
  29 static const base::FilePath::CharType kReferenceFile[] =
  30 #if defined (OS_WIN)
  31     FILE_PATH_LITERAL("human-voice-win.wav");
  32 #else
  33     FILE_PATH_LITERAL("human-voice-linux.wav");
  34 #endif
  35
  36 // The javascript will load the reference file relative to its location,
  37 // which is in /webrtc on the web server. The files we are looking for are in
  38 // webrtc/resources in the chrome/test/data folder.
  39 static const char kReferenceFileRelativeUrl[] =
  40 #if defined (OS_WIN)
  41     "resources/human-voice-win.wav";
  42 #else
  43     "resources/human-voice-linux.wav";
  44 #endif
  45
  46 static const char kMainWebrtcTestHtmlPage[] =
  47     "/webrtc/webrtc_audio_quality_test.html";
  48
  49 // Test we can set up a WebRTC call and play audio through it.
  50 //
  51 // You must have the src-internal solution in your .gclient to put the required
  52 // pyauto_private directory into chrome/test/data/.
  53 //
  54 // This test will only work on machines that have been configured to record
  55 // their own input.
  56 //
  57 // On Linux:
  58 // 1. # sudo apt-get install pavucontrol
  59 // 2. For the user who will run the test: # pavucontrol
  60 // 3. In a separate terminal, # arecord dummy
  61 // 4. In pavucontrol, go to the recording tab.
  62 // 5. For the ALSA plug-in [aplay]: ALSA Capture from, change from <x> to
  63 //    <Monitor of x>, where x is whatever your primary sound device is called.
  64 // 6. Try launching chrome as the target user on the target machine, try
  65 //    playing, say, a YouTube video, and record with # arecord -f dat tmp.dat.
  66 //    Verify the recording with aplay (should have recorded what you played
  67 //    from chrome).
  68 //
  69 // Note: the volume for ALL your input devices will be forced to 100% by
  70 //       running this test on Linux.
  71 //
  72 // On Windows 7:
  73 // 1. Control panel > Sound > Manage audio devices.
  74 // 2. In the recording tab, right-click in an empty space in the pane with the
  75 //    devices. Tick 'show disabled devices'.
  76 // 3. You should see a 'stero mix' device - this is what your speakers output.
  77 //    Right click > Properties.
  78 // 4. In the Listen tab for the mix device, check the 'listen to this device'
  79 //    checkbox. Ensure the mix device is the default recording device.
  80 // 5. Launch chrome and try playing a video with sound. You should see
  81 //    in the volume meter for the mix device. Configure the mix device to have
  82 //    50 / 100 in level. Also go into the playback tab, right-click Speakers,
  83 //    and set that level to 50 / 100. Otherwise you will get distortion in
  84 //    the recording.
  85 class WebRtcAudioQualityBrowserTest : public WebRtcTestBase,
  86                                       public testing::WithParamInterface<bool> {
  87  public:
  88   WebRtcAudioQualityBrowserTest() {}
  89   virtual void SetUpInProcessBrowserTestFixture() OVERRIDE {
  90     test::PeerConnectionServerRunner::KillAllPeerConnectionServers();
  91     DetectErrorsInJavaScript();  // Look for errors in our rather complex js.
  92   }
  93
  94   virtual void SetUpCommandLine(CommandLine* command_line) OVERRIDE {
  95     // This test expects real device handling and requires a real webcam / audio
  96     // device; it will not work with fake devices.
  97     EXPECT_FALSE(command_line->HasSwitch(
  98         switches::kUseFakeDeviceForMediaStream));
  99     EXPECT_FALSE(command_line->HasSwitch(
 100         switches::kUseFakeUIForMediaStream));
 101
 102     bool enable_audio_track_processing = GetParam();
 103     if (enable_audio_track_processing)
 104       command_line->AppendSwitch(switches::kEnableAudioTrackProcessing);
 105   }
 106
 107   void AddAudioFile(const std::string& input_file_relative_url,
 108                     content::WebContents* tab_contents) {
 109     EXPECT_EQ("ok-added", ExecuteJavascript(
 110         "addAudioFile('" + input_file_relative_url + "')", tab_contents));
 111   }
 112
 113   void PlayAudioFile(content::WebContents* tab_contents) {
 114     EXPECT_EQ("ok-playing", ExecuteJavascript("playAudioFile()", tab_contents));
 115   }
 116
 117   void EstablishCall(content::WebContents* from_tab,
 118                      content::WebContents* to_tab) {
 119     EXPECT_EQ("ok-negotiating",
 120               ExecuteJavascript("negotiateCall()", from_tab));
 121
 122     // Ensure the call gets up on both sides.
 123     EXPECT_TRUE(test::PollingWaitUntil("getPeerConnectionReadyState()",
 124                                        "active", from_tab));
 125     EXPECT_TRUE(test::PollingWaitUntil("getPeerConnectionReadyState()",
 126                                        "active", to_tab));
 127   }
 128
 129   base::FilePath CreateTemporaryWaveFile() {
 130     base::FilePath filename;
 131     EXPECT_TRUE(base::CreateTemporaryFile(&filename));
 132     base::FilePath wav_filename =
 133         filename.AddExtension(FILE_PATH_LITERAL(".wav"));
 134     EXPECT_TRUE(base::Move(filename, wav_filename));
 135     return wav_filename;
 136   }
 137
 138   test::PeerConnectionServerRunner peerconnection_server_;
 139 };
 140
 141 class AudioRecorder {
 142  public:
 143   AudioRecorder(): recording_application_(base::kNullProcessHandle) {}
 144   ~AudioRecorder() {}
 145
 146   // Starts the recording program for the specified duration. Returns true
 147   // on success.
 148   bool StartRecording(int duration_sec, const base::FilePath& output_file,
 149                       bool mono) {
 150     EXPECT_EQ(base::kNullProcessHandle, recording_application_)
 151         << "Tried to record, but is already recording.";
 152
 153     CommandLine command_line(CommandLine::NO_PROGRAM);
 154 #if defined(OS_WIN)
 155     // This disable is required to run SoundRecorder.exe on 64-bit Windows
 156     // from a 32-bit binary. We need to load the wow64 disable function from
 157     // the DLL since it doesn't exist on Windows XP.
 158     // TODO(phoglund): find some cleaner solution than using SoundRecorder.exe.
 159     base::ScopedNativeLibrary kernel32_lib(base::FilePath(L"kernel32"));
 160     if (kernel32_lib.is_valid()) {
 161       typedef BOOL (WINAPI* Wow64DisableWow64FSRedirection)(PVOID*);
 162       Wow64DisableWow64FSRedirection wow_64_disable_wow_64_fs_redirection;
 163       wow_64_disable_wow_64_fs_redirection =
 164           reinterpret_cast<Wow64DisableWow64FSRedirection>(
 165               kernel32_lib.GetFunctionPointer(
 166                   "Wow64DisableWow64FsRedirection"));
 167       if (wow_64_disable_wow_64_fs_redirection != NULL) {
 168         PVOID* ignored = NULL;
 169         wow_64_disable_wow_64_fs_redirection(ignored);
 170       }
 171     }
 172
 173     char duration_in_hms[128] = {0};
 174     struct tm duration_tm = {0};
 175     duration_tm.tm_sec = duration_sec;
 176     EXPECT_NE(0u, strftime(duration_in_hms, arraysize(duration_in_hms),
 177                            "%H:%M:%S", &duration_tm));
 178
 179     command_line.SetProgram(
 180         base::FilePath(FILE_PATH_LITERAL("SoundRecorder.exe")));
 181     command_line.AppendArg("/FILE");
 182     command_line.AppendArgPath(output_file);
 183     command_line.AppendArg("/DURATION");
 184     command_line.AppendArg(duration_in_hms);
 185 #else
 186     int num_channels = mono ? 1 : 2;
 187     command_line.SetProgram(base::FilePath("arecord"));
 188     command_line.AppendArg("-d");
 189     command_line.AppendArg(base::StringPrintf("%d", duration_sec));
 190     command_line.AppendArg("-f");
 191     command_line.AppendArg("dat");
 192     command_line.AppendArg("-c");
 193     command_line.AppendArg(base::StringPrintf("%d", num_channels));
 194     command_line.AppendArgPath(output_file);
 195 #endif
 196
 197     VLOG(0) << "Running " << command_line.GetCommandLineString();
 198     return base::LaunchProcess(command_line, base::LaunchOptions(),
 199                                &recording_application_);
 200   }
 201
 202   // Joins the recording program. Returns true on success.
 203   bool WaitForRecordingToEnd() {
 204     int exit_code = -1;
 205     base::WaitForExitCode(recording_application_, &exit_code);
 206     return exit_code == 0;
 207   }
 208  private:
 209   base::ProcessHandle recording_application_;
 210 };
 211
 212 bool ForceMicrophoneVolumeTo100Percent() {
 213 #if defined(OS_WIN)
 214   CommandLine command_line(test::GetReferenceFilesDir().Append(
 215       FILE_PATH_LITERAL("force_mic_volume_max.exe")));
 216   VLOG(0) << "Running " << command_line.GetCommandLineString();
 217   std::string result;
 218   if (!base::GetAppOutput(command_line, &result)) {
 219     LOG(ERROR) << "Failed to set source volume: output was " << result;
 220     return false;
 221   }
 222 #else
 223   // Just force the volume of, say the first 5 devices. A machine will rarely
 224   // have more input sources than that. This is way easier than finding the
 225   // input device we happen to be using.
 226   for (int device_index = 0; device_index < 5; ++device_index) {
 227     std::string result;
 228     const std::string kHundredPercentVolume = "65536";
 229     CommandLine command_line(base::FilePath(FILE_PATH_LITERAL("pacmd")));
 230     command_line.AppendArg("set-source-volume");
 231     command_line.AppendArg(base::StringPrintf("%d", device_index));
 232     command_line.AppendArg(kHundredPercentVolume);
 233     VLOG(0) << "Running " << command_line.GetCommandLineString();
 234     if (!base::GetAppOutput(command_line, &result)) {
 235       LOG(ERROR) << "Failed to set source volume: output was " << result;
 236       return false;
 237     }
 238   }
 239 #endif
 240   return true;
 241 }
 242
 243 // Removes silence from beginning and end of the |input_audio_file| and writes
 244 // the result to the |output_audio_file|. Returns true on success.
 245 bool RemoveSilence(const base::FilePath& input_file,
 246                    const base::FilePath& output_file) {
 247   // SOX documentation for silence command: http://sox.sourceforge.net/sox.html
 248   // To remove the silence from both beginning and end of the audio file, we
 249   // call sox silence command twice: once on normal file and again on its
 250   // reverse, then we reverse the final output.
 251   // Silence parameters are (in sequence):
 252   // ABOVE_PERIODS: The period for which silence occurs. Value 1 is used for
 253   //                 silence at beginning of audio.
 254   // DURATION: the amount of time in seconds that non-silence must be detected
 255   //           before sox stops trimming audio.
 256   // THRESHOLD: value used to indicate what sample value is treates as silence.
 257   const char* kAbovePeriods = "1";
 258   const char* kDuration = "2";
 259   const char* kTreshold = "5%";
 260
 261 #if defined(OS_WIN)
 262   CommandLine command_line(test::GetReferenceFilesDir().Append(
 263       FILE_PATH_LITERAL("sox.exe")));
 264 #else
 265   CommandLine command_line(base::FilePath(FILE_PATH_LITERAL("sox")));
 266 #endif
 267   command_line.AppendArgPath(input_file);
 268   command_line.AppendArgPath(output_file);
 269   command_line.AppendArg("silence");
 270   command_line.AppendArg(kAbovePeriods);
 271   command_line.AppendArg(kDuration);
 272   command_line.AppendArg(kTreshold);
 273   command_line.AppendArg("reverse");
 274   command_line.AppendArg("silence");
 275   command_line.AppendArg(kAbovePeriods);
 276   command_line.AppendArg(kDuration);
 277   command_line.AppendArg(kTreshold);
 278   command_line.AppendArg("reverse");
 279
 280   VLOG(0) << "Running " << command_line.GetCommandLineString();
 281   std::string result;
 282   bool ok = base::GetAppOutput(command_line, &result);
 283   VLOG(0) << "Output was:\n\n" << result;
 284   return ok;
 285 }
 286
 287 bool CanParseAsFloat(const std::string& value) {
 288   return atof(value.c_str()) != 0 || value == "0";
 289 }
 290
 291 // Runs PESQ to compare |reference_file| to a |actual_file|. The |sample_rate|
 292 // can be either 16000 or 8000.
 293 //
 294 // PESQ is only mono-aware, so the files should preferably be recorded in mono.
 295 // Furthermore it expects the file to be 16 rather than 32 bits, even though
 296 // 32 bits might work. The audio bandwidth of the two files should be the same
 297 // e.g. don't compare a 32 kHz file to a 8 kHz file.
 298 //
 299 // The raw score in MOS is written to |raw_mos|, whereas the MOS-LQO score is
 300 // written to mos_lqo. The scores are returned as floats in string form (e.g.
 301 // "3.145", etc). Returns true on success.
 302 bool RunPesq(const base::FilePath& reference_file,
 303              const base::FilePath& actual_file,
 304              int sample_rate, std::string* raw_mos, std::string* mos_lqo) {
 305   // PESQ will break if the paths are too long (!).
 306   EXPECT_LT(reference_file.value().length(), 128u);
 307   EXPECT_LT(actual_file.value().length(), 128u);
 308
 309 #if defined(OS_WIN)
 310   base::FilePath pesq_path =
 311       test::GetReferenceFilesDir().Append(FILE_PATH_LITERAL("pesq.exe"));
 312 #else
 313   base::FilePath pesq_path =
 314       test::GetReferenceFilesDir().Append(FILE_PATH_LITERAL("pesq"));
 315 #endif
 316
 317   if (!base::PathExists(pesq_path)) {
 318     LOG(ERROR) << "Missing PESQ binary in " << pesq_path.value();
 319     return false;
 320   }
 321
 322   CommandLine command_line(pesq_path);
 323   command_line.AppendArg(base::StringPrintf("+%d", sample_rate));
 324   command_line.AppendArgPath(reference_file);
 325   command_line.AppendArgPath(actual_file);
 326
 327   VLOG(0) << "Running " << command_line.GetCommandLineString();
 328   std::string result;
 329   if (!base::GetAppOutput(command_line, &result)) {
 330     LOG(ERROR) << "Failed to run PESQ.";
 331     return false;
 332   }
 333   VLOG(0) << "Output was:\n\n" << result;
 334
 335   const std::string result_anchor = "Prediction (Raw MOS, MOS-LQO):  = ";
 336   std::size_t anchor_pos = result.find(result_anchor);
 337   if (anchor_pos == std::string::npos) {
 338     LOG(ERROR) << "PESQ was not able to compute a score; we probably recorded "
 339         << "only silence.";
 340     return false;
 341   }
 342
 343   // There are two tab-separated numbers on the format x.xxx, e.g. 5 chars each.
 344   std::size_t first_number_pos = anchor_pos + result_anchor.length();
 345   *raw_mos = result.substr(first_number_pos, 5);
 346   EXPECT_TRUE(CanParseAsFloat(*raw_mos)) << "Failed to parse raw MOS number.";
 347   *mos_lqo = result.substr(first_number_pos + 5 + 1, 5);
 348   EXPECT_TRUE(CanParseAsFloat(*mos_lqo)) << "Failed to parse MOS LQO number.";
 349
 350   return true;
 351 }
 352
 353 static const bool kRunTestsWithFlag[] = { false, true };
 354 INSTANTIATE_TEST_CASE_P(WebRtcAudioQualityBrowserTests,
 355                         WebRtcAudioQualityBrowserTest,
 356                         testing::ValuesIn(kRunTestsWithFlag));
 357
 358 #if defined(OS_LINUX) || defined(OS_WIN)
 359 // Only implemented on Linux and Windows for now.
 360 #define MAYBE_MANUAL_TestAudioQuality MANUAL_TestAudioQuality
 361 #else
 362 #define MAYBE_MANUAL_TestAudioQuality DISABLED_MANUAL_TestAudioQuality
 363 #endif
 364
 365 IN_PROC_BROWSER_TEST_P(WebRtcAudioQualityBrowserTest,
 366                        MAYBE_MANUAL_TestAudioQuality) {
 367 #if defined(OS_WIN)
 368   if (base::win::GetVersion() < base::win::VERSION_VISTA) {
 369     // It would take work to implement this on XP; not prioritized right now.
 370     LOG(ERROR) << "This test is not implemented for Windows XP.";
 371     return;
 372   }
 373 #endif
 374   ASSERT_TRUE(test::HasReferenceFilesInCheckout());
 375   ASSERT_TRUE(embedded_test_server()->InitializeAndWaitUntilReady());
 376   ASSERT_TRUE(peerconnection_server_.Start());
 377
 378   ASSERT_TRUE(ForceMicrophoneVolumeTo100Percent());
 379
 380   ui_test_utils::NavigateToURL(
 381       browser(), embedded_test_server()->GetURL(kMainWebrtcTestHtmlPage));
 382   content::WebContents* left_tab =
 383       browser()->tab_strip_model()->GetActiveWebContents();
 384
 385   chrome::AddTabAt(browser(), GURL(), -1, true);
 386   content::WebContents* right_tab =
 387       browser()->tab_strip_model()->GetActiveWebContents();
 388   ui_test_utils::NavigateToURL(
 389       browser(), embedded_test_server()->GetURL(kMainWebrtcTestHtmlPage));
 390
 391   ConnectToPeerConnectionServer("peer 1", left_tab);
 392   ConnectToPeerConnectionServer("peer 2", right_tab);
 393
 394   EXPECT_EQ("ok-peerconnection-created",
 395             ExecuteJavascript("preparePeerConnection()", left_tab));
 396
 397   AddAudioFile(kReferenceFileRelativeUrl, left_tab);
 398
 399   EstablishCall(left_tab, right_tab);
 400
 401   // Note: the media flow isn't necessarily established on the connection just
 402   // because the ready state is ok on both sides. We sleep a bit between call
 403   // establishment and playing to avoid cutting of the beginning of the audio
 404   // file.
 405   test::SleepInJavascript(left_tab, 2000);
 406
 407   base::FilePath recording = CreateTemporaryWaveFile();
 408
 409   // Note: the sound clip is about 10 seconds: record for 15 seconds to get some
 410   // safety margins on each side.
 411   AudioRecorder recorder;
 412   static int kRecordingTimeSeconds = 15;
 413   ASSERT_TRUE(recorder.StartRecording(kRecordingTimeSeconds, recording, true));
 414
 415   PlayAudioFile(left_tab);
 416
 417   ASSERT_TRUE(recorder.WaitForRecordingToEnd());
 418   VLOG(0) << "Done recording to " << recording.value() << std::endl;
 419
 420   HangUp(left_tab);
 421   WaitUntilHangupVerified(left_tab);
 422   WaitUntilHangupVerified(right_tab);
 423
 424   base::FilePath trimmed_recording = CreateTemporaryWaveFile();
 425
 426   ASSERT_TRUE(RemoveSilence(recording, trimmed_recording));
 427   VLOG(0) << "Trimmed silence: " << trimmed_recording.value() << std::endl;
 428
 429   std::string raw_mos;
 430   std::string mos_lqo;
 431   base::FilePath reference_file_in_test_dir =
 432       test::GetReferenceFilesDir().Append(kReferenceFile);
 433   ASSERT_TRUE(RunPesq(reference_file_in_test_dir, trimmed_recording, 16000,
 434                       &raw_mos, &mos_lqo));
 435
 436   perf_test::PrintResult("audio_pesq", "", "raw_mos", raw_mos, "score", true);
 437   perf_test::PrintResult("audio_pesq", "", "mos_lqo", mos_lqo, "score", true);
 438
 439   EXPECT_TRUE(base::DeleteFile(recording, false));
 440   EXPECT_TRUE(base::DeleteFile(trimmed_recording, false));
 441
 442   ASSERT_TRUE(peerconnection_server_.Stop());
 443 }