| OLD | NEW |
| (Empty) |
| 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 <stdint.h> | |
| 8 | |
| 9 #include <algorithm> | |
| 10 #include <limits> | |
| 11 #include <memory> | |
| 12 #include <vector> | |
| 13 | |
| 14 #include "base/bind.h" | |
| 15 #include "base/bind_helpers.h" | |
| 16 #include "base/command_line.h" | |
| 17 #include "base/logging.h" | |
| 18 #include "base/memory/ptr_util.h" | |
| 19 #include "base/run_loop.h" | |
| 20 #include "base/strings/string_util.h" | |
| 21 #include "base/strings/stringprintf.h" | |
| 22 #include "base/strings/sys_string_conversions.h" | |
| 23 #include "base/time/time.h" | |
| 24 #include "build/build_config.h" | |
| 25 #include "components/audio_modem/audio_modem_switches.h" | |
| 26 #include "components/audio_modem/audio_player_impl.h" | |
| 27 #include "components/audio_modem/audio_recorder_impl.h" | |
| 28 #include "components/audio_modem/public/whispernet_client.h" | |
| 29 #include "content/public/browser/browser_thread.h" | |
| 30 #include "media/audio/audio_manager.h" | |
| 31 #include "media/audio/audio_manager_base.h" | |
| 32 #include "media/base/audio_bus.h" | |
| 33 #include "third_party/webrtc/common_audio/wav_file.h" | |
| 34 | |
| 35 namespace audio_modem { | |
| 36 | |
| 37 namespace { | |
| 38 | |
| 39 const int kMaxSamples = 10000; | |
| 40 const int kTokenTimeoutMs = 2000; | |
| 41 const int kMonoChannelCount = 1; | |
| 42 | |
| 43 // UrlSafe is defined as: | |
| 44 // '/' represented by a '_' and '+' represented by a '-' | |
| 45 // TODO(ckehoe): Move this to a central place. | |
| 46 std::string FromUrlSafe(std::string token) { | |
| 47 base::ReplaceChars(token, "-", "+", &token); | |
| 48 base::ReplaceChars(token, "_", "/", &token); | |
| 49 return token; | |
| 50 } | |
| 51 std::string ToUrlSafe(std::string token) { | |
| 52 base::ReplaceChars(token, "+", "-", &token); | |
| 53 base::ReplaceChars(token, "/", "_", &token); | |
| 54 return token; | |
| 55 } | |
| 56 | |
| 57 // TODO(ckehoe): Move this to a central place. | |
| 58 std::string AudioTypeToString(AudioType audio_type) { | |
| 59 if (audio_type == AUDIBLE) | |
| 60 return "audible"; | |
| 61 if (audio_type == INAUDIBLE) | |
| 62 return "inaudible"; | |
| 63 | |
| 64 NOTREACHED() << "Got unexpected token type " << audio_type; | |
| 65 return std::string(); | |
| 66 } | |
| 67 | |
| 68 bool ReadBooleanFlag(const std::string& flag, bool default_value) { | |
| 69 const std::string flag_value = base::ToLowerASCII( | |
| 70 base::CommandLine::ForCurrentProcess()->GetSwitchValueASCII(flag)); | |
| 71 if (flag_value == "true" || flag_value == "1") | |
| 72 return true; | |
| 73 if (flag_value == "false" || flag_value == "0") | |
| 74 return false; | |
| 75 LOG_IF(ERROR, !flag_value.empty()) | |
| 76 << "Unrecognized value \"" << flag_value << " for flag " | |
| 77 << flag << ". Defaulting to " << default_value; | |
| 78 return default_value; | |
| 79 } | |
| 80 | |
| 81 } // namespace | |
| 82 | |
| 83 | |
| 84 // Public functions. | |
| 85 | |
| 86 ModemImpl::ModemImpl() : client_(nullptr), recorder_(nullptr) { | |
| 87 // TODO(rkc): Move all of these into initializer lists once it is allowed. | |
| 88 should_be_playing_[AUDIBLE] = false; | |
| 89 should_be_playing_[INAUDIBLE] = false; | |
| 90 should_be_recording_[AUDIBLE] = false; | |
| 91 should_be_recording_[INAUDIBLE] = false; | |
| 92 | |
| 93 player_enabled_[AUDIBLE] = ReadBooleanFlag( | |
| 94 switches::kAudioModemEnableAudibleBroadcast, true); | |
| 95 player_enabled_[INAUDIBLE] = ReadBooleanFlag( | |
| 96 switches::kAudioModemEnableInaudibleBroadcast, true); | |
| 97 player_[AUDIBLE] = nullptr; | |
| 98 player_[INAUDIBLE] = nullptr; | |
| 99 | |
| 100 samples_caches_.resize(2); | |
| 101 samples_caches_[AUDIBLE] = new SamplesMap(kMaxSamples); | |
| 102 samples_caches_[INAUDIBLE] = new SamplesMap(kMaxSamples); | |
| 103 } | |
| 104 | |
| 105 void ModemImpl::Initialize(WhispernetClient* client, | |
| 106 const TokensCallback& tokens_cb) { | |
| 107 DCHECK(client); | |
| 108 client_ = client; | |
| 109 tokens_cb_ = tokens_cb; | |
| 110 | |
| 111 // These will be unregistered on destruction, so unretained is safe to use. | |
| 112 client_->RegisterTokensCallback( | |
| 113 base::Bind(&ModemImpl::OnTokensFound, base::Unretained(this))); | |
| 114 client_->RegisterSamplesCallback( | |
| 115 base::Bind(&ModemImpl::OnTokenEncoded, base::Unretained(this))); | |
| 116 | |
| 117 if (!player_[AUDIBLE]) | |
| 118 player_[AUDIBLE] = new AudioPlayerImpl(); | |
| 119 player_[AUDIBLE]->Initialize(); | |
| 120 | |
| 121 if (!player_[INAUDIBLE]) | |
| 122 player_[INAUDIBLE] = new AudioPlayerImpl(); | |
| 123 player_[INAUDIBLE]->Initialize(); | |
| 124 | |
| 125 decode_cancelable_cb_.Reset(base::Bind( | |
| 126 &ModemImpl::DecodeSamplesConnector, base::Unretained(this))); | |
| 127 if (!recorder_) | |
| 128 recorder_ = new AudioRecorderImpl(); | |
| 129 recorder_->Initialize(decode_cancelable_cb_.callback()); | |
| 130 | |
| 131 dump_tokens_dir_ = base::FilePath(base::CommandLine::ForCurrentProcess() | |
| 132 ->GetSwitchValueNative(switches::kAudioModemDumpTokensToDir)); | |
| 133 } | |
| 134 | |
| 135 ModemImpl::~ModemImpl() { | |
| 136 if (player_[AUDIBLE]) | |
| 137 player_[AUDIBLE]->Finalize(); | |
| 138 if (player_[INAUDIBLE]) | |
| 139 player_[INAUDIBLE]->Finalize(); | |
| 140 if (recorder_) | |
| 141 recorder_->Finalize(); | |
| 142 | |
| 143 // Whispernet initialization may never have completed. | |
| 144 if (client_) { | |
| 145 client_->RegisterTokensCallback(TokensCallback()); | |
| 146 client_->RegisterSamplesCallback(SamplesCallback()); | |
| 147 } | |
| 148 } | |
| 149 | |
| 150 void ModemImpl::StartPlaying(AudioType type) { | |
| 151 DCHECK(type == AUDIBLE || type == INAUDIBLE); | |
| 152 should_be_playing_[type] = true; | |
| 153 // If we don't have our token encoded yet, this check will be false, for now. | |
| 154 // Once our token is encoded, OnTokenEncoded will call UpdateToken, which | |
| 155 // will call this code again (if we're still supposed to be playing). | |
| 156 SamplesMap::iterator samples = | |
| 157 samples_caches_[type]->Get(playing_token_[type]); | |
| 158 if (samples != samples_caches_[type]->end()) { | |
| 159 DCHECK(!playing_token_[type].empty()); | |
| 160 if (player_enabled_[type]) { | |
| 161 started_playing_[type] = base::Time::Now(); | |
| 162 player_[type]->Play(samples->second); | |
| 163 | |
| 164 // If we're playing, we always record to hear what we are playing. | |
| 165 recorder_->Record(); | |
| 166 } else { | |
| 167 DVLOG(3) << "Skipping playback for disabled " << AudioTypeToString(type) | |
| 168 << " player."; | |
| 169 } | |
| 170 } | |
| 171 } | |
| 172 | |
| 173 void ModemImpl::StopPlaying(AudioType type) { | |
| 174 DCHECK(type == AUDIBLE || type == INAUDIBLE); | |
| 175 should_be_playing_[type] = false; | |
| 176 player_[type]->Stop(); | |
| 177 // If we were only recording to hear our own played tokens, stop. | |
| 178 if (!should_be_recording_[AUDIBLE] && !should_be_recording_[INAUDIBLE]) | |
| 179 recorder_->Stop(); | |
| 180 playing_token_[type] = std::string(); | |
| 181 } | |
| 182 | |
| 183 void ModemImpl::StartRecording(AudioType type) { | |
| 184 DCHECK(type == AUDIBLE || type == INAUDIBLE); | |
| 185 should_be_recording_[type] = true; | |
| 186 recorder_->Record(); | |
| 187 } | |
| 188 | |
| 189 void ModemImpl::StopRecording(AudioType type) { | |
| 190 DCHECK(type == AUDIBLE || type == INAUDIBLE); | |
| 191 should_be_recording_[type] = false; | |
| 192 recorder_->Stop(); | |
| 193 } | |
| 194 | |
| 195 void ModemImpl::SetToken(AudioType type, | |
| 196 const std::string& url_safe_token) { | |
| 197 DCHECK(type == AUDIBLE || type == INAUDIBLE); | |
| 198 std::string token = FromUrlSafe(url_safe_token); | |
| 199 if (samples_caches_[type]->Get(token) == samples_caches_[type]->end()) { | |
| 200 client_->EncodeToken(token, type, token_params_); | |
| 201 } else { | |
| 202 UpdateToken(type, token); | |
| 203 } | |
| 204 } | |
| 205 | |
| 206 const std::string ModemImpl::GetToken(AudioType type) const { | |
| 207 return playing_token_[type]; | |
| 208 } | |
| 209 | |
| 210 bool ModemImpl::IsPlayingTokenHeard(AudioType type) const { | |
| 211 base::TimeDelta tokenTimeout = | |
| 212 base::TimeDelta::FromMilliseconds(kTokenTimeoutMs); | |
| 213 | |
| 214 // This is a bit of a hack. If we haven't been playing long enough, | |
| 215 // return true to avoid tripping an audio fail alarm. | |
| 216 if (base::Time::Now() - started_playing_[type] < tokenTimeout) | |
| 217 return true; | |
| 218 | |
| 219 return base::Time::Now() - heard_own_token_[type] < tokenTimeout; | |
| 220 } | |
| 221 | |
| 222 void ModemImpl::SetTokenParams(AudioType type, const TokenParameters& params) { | |
| 223 DCHECK_GT(params.length, 0u); | |
| 224 token_params_[type] = params; | |
| 225 | |
| 226 // TODO(ckehoe): Make whispernet handle different token lengths | |
| 227 // simultaneously without reinitializing the decoder over and over. | |
| 228 } | |
| 229 | |
| 230 // static | |
| 231 std::unique_ptr<Modem> Modem::Create() { | |
| 232 return base::WrapUnique<Modem>(new ModemImpl); | |
| 233 } | |
| 234 | |
| 235 // Private functions. | |
| 236 | |
| 237 void ModemImpl::OnTokenEncoded( | |
| 238 AudioType type, | |
| 239 const std::string& token, | |
| 240 const scoped_refptr<media::AudioBusRefCounted>& samples) { | |
| 241 samples_caches_[type]->Put(token, samples); | |
| 242 DumpToken(type, token, samples.get()); | |
| 243 UpdateToken(type, token); | |
| 244 } | |
| 245 | |
| 246 void ModemImpl::OnTokensFound(const std::vector<AudioToken>& tokens) { | |
| 247 std::vector<AudioToken> tokens_to_report; | |
| 248 for (const auto& token : tokens) { | |
| 249 AudioType type = token.audible ? AUDIBLE : INAUDIBLE; | |
| 250 if (playing_token_[type] == token.token) | |
| 251 heard_own_token_[type] = base::Time::Now(); | |
| 252 | |
| 253 if (should_be_recording_[AUDIBLE] && token.audible) { | |
| 254 tokens_to_report.push_back(token); | |
| 255 } else if (should_be_recording_[INAUDIBLE] && !token.audible) { | |
| 256 tokens_to_report.push_back(token); | |
| 257 } | |
| 258 } | |
| 259 | |
| 260 if (!tokens_to_report.empty()) | |
| 261 tokens_cb_.Run(tokens_to_report); | |
| 262 } | |
| 263 | |
| 264 void ModemImpl::UpdateToken(AudioType type, const std::string& token) { | |
| 265 DCHECK(type == AUDIBLE || type == INAUDIBLE); | |
| 266 if (playing_token_[type] == token) | |
| 267 return; | |
| 268 | |
| 269 // Update token. | |
| 270 playing_token_[type] = token; | |
| 271 | |
| 272 // If we are supposed to be playing this token type at this moment, switch | |
| 273 // out playback with the new samples. | |
| 274 if (should_be_playing_[type]) | |
| 275 RestartPlaying(type); | |
| 276 } | |
| 277 | |
| 278 void ModemImpl::RestartPlaying(AudioType type) { | |
| 279 DCHECK(type == AUDIBLE || type == INAUDIBLE); | |
| 280 // We should already have this token in the cache. This function is not | |
| 281 // called from anywhere except update token and only once we have our samples | |
| 282 // in the cache. | |
| 283 DCHECK(samples_caches_[type]->Get(playing_token_[type]) != | |
| 284 samples_caches_[type]->end()); | |
| 285 | |
| 286 player_[type]->Stop(); | |
| 287 StartPlaying(type); | |
| 288 } | |
| 289 | |
| 290 void ModemImpl::DecodeSamplesConnector(const std::string& samples) { | |
| 291 // If we are either supposed to be recording *or* playing, audible or | |
| 292 // inaudible, we should be decoding that type. This is so that if we are | |
| 293 // just playing, we will still decode our recorded token so we can check | |
| 294 // if we heard our own token. Whether or not we report the token to the | |
| 295 // server is checked for and handled in OnTokensFound. | |
| 296 | |
| 297 bool decode_audible = | |
| 298 should_be_recording_[AUDIBLE] || should_be_playing_[AUDIBLE]; | |
| 299 bool decode_inaudible = | |
| 300 should_be_recording_[INAUDIBLE] || should_be_playing_[INAUDIBLE]; | |
| 301 | |
| 302 if (decode_audible && decode_inaudible) { | |
| 303 client_->DecodeSamples(BOTH, samples, token_params_); | |
| 304 } else if (decode_audible) { | |
| 305 client_->DecodeSamples(AUDIBLE, samples, token_params_); | |
| 306 } else if (decode_inaudible) { | |
| 307 client_->DecodeSamples(INAUDIBLE, samples, token_params_); | |
| 308 } | |
| 309 } | |
| 310 | |
| 311 void ModemImpl::DumpToken(AudioType audio_type, | |
| 312 const std::string& token, | |
| 313 const media::AudioBus* samples) { | |
| 314 if (dump_tokens_dir_.empty()) | |
| 315 return; | |
| 316 | |
| 317 // Convert the samples to 16-bit integers. | |
| 318 std::vector<int16_t> int_samples; | |
| 319 int_samples.reserve(samples->frames()); | |
| 320 for (int i = 0; i < samples->frames(); i++) { | |
| 321 int_samples.push_back(round( | |
| 322 samples->channel(0)[i] * std::numeric_limits<int16_t>::max())); | |
| 323 } | |
| 324 DCHECK_EQ(static_cast<int>(int_samples.size()), samples->frames()); | |
| 325 DCHECK_EQ(kMonoChannelCount, samples->channels()); | |
| 326 | |
| 327 const std::string filename = base::StringPrintf("%s %s.wav", | |
| 328 AudioTypeToString(audio_type).c_str(), ToUrlSafe(token).c_str()); | |
| 329 DVLOG(3) << "Dumping token " << filename; | |
| 330 | |
| 331 std::string file_str; | |
| 332 #if defined(OS_WIN) | |
| 333 base::FilePath file_path = dump_tokens_dir_.Append( | |
| 334 base::SysNativeMBToWide(filename)); | |
| 335 file_str = base::SysWideToNativeMB(file_path.value()); | |
| 336 #else | |
| 337 file_str = dump_tokens_dir_.Append(filename).value(); | |
| 338 #endif | |
| 339 | |
| 340 webrtc::WavWriter writer(file_str, kDefaultSampleRate, kMonoChannelCount); | |
| 341 writer.WriteSamples(int_samples.data(), int_samples.size()); | |
| 342 } | |
| 343 | |
| 344 } // namespace audio_modem | |
| OLD | NEW |