components/audio_modem/modem_impl.cc

   1 // Copyright 2015 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 "components/audio_modem/modem_impl.h"
   6
   7 #include <algorithm>
   8 #include <limits>
   9 #include <vector>
  10
  11 #include "base/bind.h"
  12 #include "base/bind_helpers.h"
  13 #include "base/command_line.h"
  14 #include "base/logging.h"
  15 #include "base/memory/scoped_ptr.h"
  16 #include "base/run_loop.h"
  17 #include "base/strings/string_util.h"
  18 #include "base/strings/stringprintf.h"
  19 #include "base/strings/sys_string_conversions.h"
  20 #include "base/time/time.h"
  21 #include "components/audio_modem/audio_modem_switches.h"
  22 #include "components/audio_modem/audio_player_impl.h"
  23 #include "components/audio_modem/audio_recorder_impl.h"
  24 #include "components/audio_modem/public/whispernet_client.h"
  25 #include "content/public/browser/browser_thread.h"
  26 #include "media/audio/audio_manager.h"
  27 #include "media/audio/audio_manager_base.h"
  28 #include "media/base/audio_bus.h"
  29 #include "third_party/webrtc/common_audio/wav_file.h"
  30
  31 namespace audio_modem {
  32
  33 namespace {
  34
  35 const int kMaxSamples = 10000;
  36 const int kTokenTimeoutMs = 2000;
  37 const int kMonoChannelCount = 1;
  38
  39 // UrlSafe is defined as:
  40 // '/' represented by a '_' and '+' represented by a '-'
  41 // TODO(ckehoe): Move this to a central place.
  42 std::string FromUrlSafe(std::string token) {
  43   base::ReplaceChars(token, "-", "+", &token);
  44   base::ReplaceChars(token, "_", "/", &token);
  45   return token;
  46 }
  47 std::string ToUrlSafe(std::string token) {
  48   base::ReplaceChars(token, "+", "-", &token);
  49   base::ReplaceChars(token, "/", "_", &token);
  50   return token;
  51 }
  52
  53 // TODO(ckehoe): Move this to a central place.
  54 std::string AudioTypeToString(AudioType audio_type) {
  55   if (audio_type == AUDIBLE)
  56     return "audible";
  57   if (audio_type == INAUDIBLE)
  58     return "inaudible";
  59
  60   NOTREACHED() << "Got unexpected token type " << audio_type;
  61   return std::string();
  62 }
  63
  64 bool ReadBooleanFlag(const std::string& flag, bool default_value) {
  65   const std::string flag_value = base::ToLowerASCII(
  66       base::CommandLine::ForCurrentProcess()->GetSwitchValueASCII(flag));
  67   if (flag_value == "true" || flag_value == "1")
  68     return true;
  69   if (flag_value == "false" || flag_value == "0")
  70     return false;
  71   LOG_IF(ERROR, !flag_value.empty())
  72       << "Unrecognized value \"" << flag_value << " for flag "
  73       << flag << ". Defaulting to " << default_value;
  74   return default_value;
  75 }
  76
  77 }  // namespace
  78
  79
  80 // Public functions.
  81
  82 ModemImpl::ModemImpl() : client_(nullptr), recorder_(nullptr) {
  83   // TODO(rkc): Move all of these into initializer lists once it is allowed.
  84   should_be_playing_[AUDIBLE] = false;
  85   should_be_playing_[INAUDIBLE] = false;
  86   should_be_recording_[AUDIBLE] = false;
  87   should_be_recording_[INAUDIBLE] = false;
  88
  89   player_enabled_[AUDIBLE] = ReadBooleanFlag(
  90       switches::kAudioModemEnableAudibleBroadcast, true);
  91   player_enabled_[INAUDIBLE] = ReadBooleanFlag(
  92       switches::kAudioModemEnableInaudibleBroadcast, true);
  93   player_[AUDIBLE] = nullptr;
  94   player_[INAUDIBLE] = nullptr;
  95
  96   samples_caches_.resize(2);
  97   samples_caches_[AUDIBLE] = new SamplesMap(kMaxSamples);
  98   samples_caches_[INAUDIBLE] = new SamplesMap(kMaxSamples);
  99 }
 100
 101 void ModemImpl::Initialize(WhispernetClient* client,
 102                            const TokensCallback& tokens_cb) {
 103   DCHECK(client);
 104   client_ = client;
 105   tokens_cb_ = tokens_cb;
 106
 107   // These will be unregistered on destruction, so unretained is safe to use.
 108   client_->RegisterTokensCallback(
 109       base::Bind(&ModemImpl::OnTokensFound, base::Unretained(this)));
 110   client_->RegisterSamplesCallback(
 111       base::Bind(&ModemImpl::OnTokenEncoded, base::Unretained(this)));
 112
 113   if (!player_[AUDIBLE])
 114     player_[AUDIBLE] = new AudioPlayerImpl();
 115   player_[AUDIBLE]->Initialize();
 116
 117   if (!player_[INAUDIBLE])
 118     player_[INAUDIBLE] = new AudioPlayerImpl();
 119   player_[INAUDIBLE]->Initialize();
 120
 121   decode_cancelable_cb_.Reset(base::Bind(
 122       &ModemImpl::DecodeSamplesConnector, base::Unretained(this)));
 123   if (!recorder_)
 124     recorder_ = new AudioRecorderImpl();
 125   recorder_->Initialize(decode_cancelable_cb_.callback());
 126
 127   dump_tokens_dir_ = base::FilePath(base::CommandLine::ForCurrentProcess()
 128       ->GetSwitchValueNative(switches::kAudioModemDumpTokensToDir));
 129 }
 130
 131 ModemImpl::~ModemImpl() {
 132   if (player_[AUDIBLE])
 133     player_[AUDIBLE]->Finalize();
 134   if (player_[INAUDIBLE])
 135     player_[INAUDIBLE]->Finalize();
 136   if (recorder_)
 137     recorder_->Finalize();
 138
 139   // Whispernet initialization may never have completed.
 140   if (client_) {
 141     client_->RegisterTokensCallback(TokensCallback());
 142     client_->RegisterSamplesCallback(SamplesCallback());
 143   }
 144 }
 145
 146 void ModemImpl::StartPlaying(AudioType type) {
 147   DCHECK(type == AUDIBLE || type == INAUDIBLE);
 148   should_be_playing_[type] = true;
 149   // If we don't have our token encoded yet, this check will be false, for now.
 150   // Once our token is encoded, OnTokenEncoded will call UpdateToken, which
 151   // will call this code again (if we're still supposed to be playing).
 152   SamplesMap::iterator samples =
 153       samples_caches_[type]->Get(playing_token_[type]);
 154   if (samples != samples_caches_[type]->end()) {
 155     DCHECK(!playing_token_[type].empty());
 156     if (player_enabled_[type]) {
 157       started_playing_[type] = base::Time::Now();
 158       player_[type]->Play(samples->second);
 159
 160       // If we're playing, we always record to hear what we are playing.
 161       recorder_->Record();
 162     } else {
 163       DVLOG(3) << "Skipping playback for disabled " << AudioTypeToString(type)
 164                << " player.";
 165     }
 166   }
 167 }
 168
 169 void ModemImpl::StopPlaying(AudioType type) {
 170   DCHECK(type == AUDIBLE || type == INAUDIBLE);
 171   should_be_playing_[type] = false;
 172   player_[type]->Stop();
 173   // If we were only recording to hear our own played tokens, stop.
 174   if (!should_be_recording_[AUDIBLE] && !should_be_recording_[INAUDIBLE])
 175     recorder_->Stop();
 176   playing_token_[type] = std::string();
 177 }
 178
 179 void ModemImpl::StartRecording(AudioType type) {
 180   DCHECK(type == AUDIBLE || type == INAUDIBLE);
 181   should_be_recording_[type] = true;
 182   recorder_->Record();
 183 }
 184
 185 void ModemImpl::StopRecording(AudioType type) {
 186   DCHECK(type == AUDIBLE || type == INAUDIBLE);
 187   should_be_recording_[type] = false;
 188   recorder_->Stop();
 189 }
 190
 191 void ModemImpl::SetToken(AudioType type,
 192                          const std::string& url_safe_token) {
 193   DCHECK(type == AUDIBLE || type == INAUDIBLE);
 194   std::string token = FromUrlSafe(url_safe_token);
 195   if (samples_caches_[type]->Get(token) == samples_caches_[type]->end()) {
 196     client_->EncodeToken(token, type, token_params_);
 197   } else {
 198     UpdateToken(type, token);
 199   }
 200 }
 201
 202 const std::string ModemImpl::GetToken(AudioType type) const {
 203   return playing_token_[type];
 204 }
 205
 206 bool ModemImpl::IsPlayingTokenHeard(AudioType type) const {
 207   base::TimeDelta tokenTimeout =
 208       base::TimeDelta::FromMilliseconds(kTokenTimeoutMs);
 209
 210   // This is a bit of a hack. If we haven't been playing long enough,
 211   // return true to avoid tripping an audio fail alarm.
 212   if (base::Time::Now() - started_playing_[type] < tokenTimeout)
 213     return true;
 214
 215   return base::Time::Now() - heard_own_token_[type] < tokenTimeout;
 216 }
 217
 218 void ModemImpl::SetTokenParams(AudioType type, const TokenParameters& params) {
 219   DCHECK_GT(params.length, 0u);
 220   token_params_[type] = params;
 221
 222   // TODO(ckehoe): Make whispernet handle different token lengths
 223   // simultaneously without reinitializing the decoder over and over.
 224 }
 225
 226 // static
 227 scoped_ptr<Modem> Modem::Create() {
 228   return make_scoped_ptr<Modem>(new ModemImpl);
 229 }
 230
 231 // Private functions.
 232
 233 void ModemImpl::OnTokenEncoded(
 234     AudioType type,
 235     const std::string& token,
 236     const scoped_refptr<media::AudioBusRefCounted>& samples) {
 237   samples_caches_[type]->Put(token, samples);
 238   DumpToken(type, token, samples.get());
 239   UpdateToken(type, token);
 240 }
 241
 242 void ModemImpl::OnTokensFound(const std::vector<AudioToken>& tokens) {
 243   std::vector<AudioToken> tokens_to_report;
 244   for (const auto& token : tokens) {
 245     AudioType type = token.audible ? AUDIBLE : INAUDIBLE;
 246     if (playing_token_[type] == token.token)
 247       heard_own_token_[type] = base::Time::Now();
 248
 249     if (should_be_recording_[AUDIBLE] && token.audible) {
 250       tokens_to_report.push_back(token);
 251     } else if (should_be_recording_[INAUDIBLE] && !token.audible) {
 252       tokens_to_report.push_back(token);
 253     }
 254   }
 255
 256   if (!tokens_to_report.empty())
 257     tokens_cb_.Run(tokens_to_report);
 258 }
 259
 260 void ModemImpl::UpdateToken(AudioType type, const std::string& token) {
 261   DCHECK(type == AUDIBLE || type == INAUDIBLE);
 262   if (playing_token_[type] == token)
 263     return;
 264
 265   // Update token.
 266   playing_token_[type] = token;
 267
 268   // If we are supposed to be playing this token type at this moment, switch
 269   // out playback with the new samples.
 270   if (should_be_playing_[type])
 271     RestartPlaying(type);
 272 }
 273
 274 void ModemImpl::RestartPlaying(AudioType type) {
 275   DCHECK(type == AUDIBLE || type == INAUDIBLE);
 276   // We should already have this token in the cache. This function is not
 277   // called from anywhere except update token and only once we have our samples
 278   // in the cache.
 279   DCHECK(samples_caches_[type]->Get(playing_token_[type]) !=
 280          samples_caches_[type]->end());
 281
 282   player_[type]->Stop();
 283   StartPlaying(type);
 284 }
 285
 286 void ModemImpl::DecodeSamplesConnector(const std::string& samples) {
 287   // If we are either supposed to be recording *or* playing, audible or
 288   // inaudible, we should be decoding that type. This is so that if we are
 289   // just playing, we will still decode our recorded token so we can check
 290   // if we heard our own token. Whether or not we report the token to the
 291   // server is checked for and handled in OnTokensFound.
 292
 293   bool decode_audible =
 294       should_be_recording_[AUDIBLE] || should_be_playing_[AUDIBLE];
 295   bool decode_inaudible =
 296       should_be_recording_[INAUDIBLE] || should_be_playing_[INAUDIBLE];
 297
 298   if (decode_audible && decode_inaudible) {
 299     client_->DecodeSamples(BOTH, samples, token_params_);
 300   } else if (decode_audible) {
 301     client_->DecodeSamples(AUDIBLE, samples, token_params_);
 302   } else if (decode_inaudible) {
 303     client_->DecodeSamples(INAUDIBLE, samples, token_params_);
 304   }
 305 }
 306
 307 void ModemImpl::DumpToken(AudioType audio_type,
 308                           const std::string& token,
 309                           const media::AudioBus* samples) {
 310   if (dump_tokens_dir_.empty())
 311     return;
 312
 313   // Convert the samples to 16-bit integers.
 314   std::vector<int16_t> int_samples;
 315   int_samples.reserve(samples->frames());
 316   for (int i = 0; i < samples->frames(); i++) {
 317     int_samples.push_back(round(
 318         samples->channel(0)[i] * std::numeric_limits<int16_t>::max()));
 319   }
 320   DCHECK_EQ(static_cast<int>(int_samples.size()), samples->frames());
 321   DCHECK_EQ(kMonoChannelCount, samples->channels());
 322
 323   const std::string filename = base::StringPrintf("%s %s.wav",
 324       AudioTypeToString(audio_type).c_str(), ToUrlSafe(token).c_str());
 325   DVLOG(3) << "Dumping token " << filename;
 326
 327   std::string file_str;
 328 #if defined(OS_WIN)
 329   base::FilePath file_path = dump_tokens_dir_.Append(
 330       base::SysNativeMBToWide(filename));
 331   file_str = base::SysWideToNativeMB(file_path.value());
 332 #else
 333   file_str = dump_tokens_dir_.Append(filename).value();
 334 #endif
 335
 336   webrtc::WavWriter writer(file_str, kDefaultSampleRate, kMonoChannelCount);
 337   writer.WriteSamples(int_samples.data(), int_samples.size());
 338 }
 339
 340 }  // namespace audio_modem