1 /* 2 * Copyright (c) 2012 The WebRTC project authors. All Rights Reserved. 3 * 4 * Use of this source code is governed by a BSD-style license 5 * that can be found in the LICENSE file in the root of the source 6 * tree. An additional intellectual property rights grant can be found 7 * in the file PATENTS. All contributing project authors may 8 * be found in the AUTHORS file in the root of the source tree. 9 */ 10 11 #ifndef WEBRTC_MODULES_AUDIO_CODING_NETEQ_NETEQ_IMPL_H_ 12 #define WEBRTC_MODULES_AUDIO_CODING_NETEQ_NETEQ_IMPL_H_ 13 14 #include <string> 15 16 #include "webrtc/base/constructormagic.h" 17 #include "webrtc/base/scoped_ptr.h" 18 #include "webrtc/base/thread_annotations.h" 19 #include "webrtc/modules/audio_coding/neteq/audio_multi_vector.h" 20 #include "webrtc/modules/audio_coding/neteq/defines.h" 21 #include "webrtc/modules/audio_coding/neteq/include/neteq.h" 22 #include "webrtc/modules/audio_coding/neteq/packet.h" // Declare PacketList. 23 #include "webrtc/modules/audio_coding/neteq/random_vector.h" 24 #include "webrtc/modules/audio_coding/neteq/rtcp.h" 25 #include "webrtc/modules/audio_coding/neteq/statistics_calculator.h" 26 #include "webrtc/typedefs.h" 27 28 namespace webrtc { 29 30 // Forward declarations. 31 class Accelerate; 32 class BackgroundNoise; 33 class BufferLevelFilter; 34 class ComfortNoise; 35 class CriticalSectionWrapper; 36 class DecisionLogic; 37 class DecoderDatabase; 38 class DelayManager; 39 class DelayPeakDetector; 40 class DtmfBuffer; 41 class DtmfToneGenerator; 42 class Expand; 43 class Merge; 44 class Nack; 45 class Normal; 46 class PacketBuffer; 47 class PayloadSplitter; 48 class PostDecodeVad; 49 class PreemptiveExpand; 50 class RandomVector; 51 class SyncBuffer; 52 class TimestampScaler; 53 struct AccelerateFactory; 54 struct DtmfEvent; 55 struct ExpandFactory; 56 struct PreemptiveExpandFactory; 57 58 class NetEqImpl : public webrtc::NetEq { 59 public: 60 // Creates a new NetEqImpl object. The object will assume ownership of all 61 // injected dependencies, and will delete them when done. 62 NetEqImpl(const NetEq::Config& config, 63 BufferLevelFilter* buffer_level_filter, 64 DecoderDatabase* decoder_database, 65 DelayManager* delay_manager, 66 DelayPeakDetector* delay_peak_detector, 67 DtmfBuffer* dtmf_buffer, 68 DtmfToneGenerator* dtmf_tone_generator, 69 PacketBuffer* packet_buffer, 70 PayloadSplitter* payload_splitter, 71 TimestampScaler* timestamp_scaler, 72 AccelerateFactory* accelerate_factory, 73 ExpandFactory* expand_factory, 74 PreemptiveExpandFactory* preemptive_expand_factory, 75 bool create_components = true); 76 77 ~NetEqImpl() override; 78 79 // Inserts a new packet into NetEq. The |receive_timestamp| is an indication 80 // of the time when the packet was received, and should be measured with 81 // the same tick rate as the RTP timestamp of the current payload. 82 // Returns 0 on success, -1 on failure. 83 int InsertPacket(const WebRtcRTPHeader& rtp_header, 84 rtc::ArrayView<const uint8_t> payload, 85 uint32_t receive_timestamp) override; 86 87 // Inserts a sync-packet into packet queue. Sync-packets are decoded to 88 // silence and are intended to keep AV-sync intact in an event of long packet 89 // losses when Video NACK is enabled but Audio NACK is not. Clients of NetEq 90 // might insert sync-packet when they observe that buffer level of NetEq is 91 // decreasing below a certain threshold, defined by the application. 92 // Sync-packets should have the same payload type as the last audio payload 93 // type, i.e. they cannot have DTMF or CNG payload type, nor a codec change 94 // can be implied by inserting a sync-packet. 95 // Returns kOk on success, kFail on failure. 96 int InsertSyncPacket(const WebRtcRTPHeader& rtp_header, 97 uint32_t receive_timestamp) override; 98 99 // Instructs NetEq to deliver 10 ms of audio data. The data is written to 100 // |output_audio|, which can hold (at least) |max_length| elements. 101 // The number of channels that were written to the output is provided in 102 // the output variable |num_channels|, and each channel contains 103 // |samples_per_channel| elements. If more than one channel is written, 104 // the samples are interleaved. 105 // The speech type is written to |type|, if |type| is not NULL. 106 // Returns kOK on success, or kFail in case of an error. 107 int GetAudio(size_t max_length, 108 int16_t* output_audio, 109 size_t* samples_per_channel, 110 size_t* num_channels, 111 NetEqOutputType* type) override; 112 113 int RegisterPayloadType(NetEqDecoder codec, 114 const std::string& codec_name, 115 uint8_t rtp_payload_type) override; 116 117 int RegisterExternalDecoder(AudioDecoder* decoder, 118 NetEqDecoder codec, 119 const std::string& codec_name, 120 uint8_t rtp_payload_type, 121 int sample_rate_hz) override; 122 123 // Removes |rtp_payload_type| from the codec database. Returns 0 on success, 124 // -1 on failure. 125 int RemovePayloadType(uint8_t rtp_payload_type) override; 126 127 bool SetMinimumDelay(int delay_ms) override; 128 129 bool SetMaximumDelay(int delay_ms) override; 130 131 int LeastRequiredDelayMs() const override; 132 133 int SetTargetDelay() override; 134 135 int TargetDelay() override; 136 137 int CurrentDelayMs() const override; 138 139 // Sets the playout mode to |mode|. 140 // Deprecated. 141 // TODO(henrik.lundin) Delete. 142 void SetPlayoutMode(NetEqPlayoutMode mode) override; 143 144 // Returns the current playout mode. 145 // Deprecated. 146 // TODO(henrik.lundin) Delete. 147 NetEqPlayoutMode PlayoutMode() const override; 148 149 // Writes the current network statistics to |stats|. The statistics are reset 150 // after the call. 151 int NetworkStatistics(NetEqNetworkStatistics* stats) override; 152 153 // Writes the current RTCP statistics to |stats|. The statistics are reset 154 // and a new report period is started with the call. 155 void GetRtcpStatistics(RtcpStatistics* stats) override; 156 157 // Same as RtcpStatistics(), but does not reset anything. 158 void GetRtcpStatisticsNoReset(RtcpStatistics* stats) override; 159 160 // Enables post-decode VAD. When enabled, GetAudio() will return 161 // kOutputVADPassive when the signal contains no speech. 162 void EnableVad() override; 163 164 // Disables post-decode VAD. 165 void DisableVad() override; 166 167 bool GetPlayoutTimestamp(uint32_t* timestamp) override; 168 169 int last_output_sample_rate_hz() const override; 170 171 int SetTargetNumberOfChannels() override; 172 173 int SetTargetSampleRate() override; 174 175 // Returns the error code for the last occurred error. If no error has 176 // occurred, 0 is returned. 177 int LastError() const override; 178 179 // Returns the error code last returned by a decoder (audio or comfort noise). 180 // When LastError() returns kDecoderErrorCode or kComfortNoiseErrorCode, check 181 // this method to get the decoder's error code. 182 int LastDecoderError() override; 183 184 // Flushes both the packet buffer and the sync buffer. 185 void FlushBuffers() override; 186 187 void PacketBufferStatistics(int* current_num_packets, 188 int* max_num_packets) const override; 189 190 void EnableNack(size_t max_nack_list_size) override; 191 192 void DisableNack() override; 193 194 std::vector<uint16_t> GetNackList(int64_t round_trip_time_ms) const override; 195 196 // This accessor method is only intended for testing purposes. 197 const SyncBuffer* sync_buffer_for_test() const; 198 199 protected: 200 static const int kOutputSizeMs = 10; 201 static const size_t kMaxFrameSize = 2880; // 60 ms @ 48 kHz. 202 // TODO(hlundin): Provide a better value for kSyncBufferSize. 203 static const size_t kSyncBufferSize = 2 * kMaxFrameSize; 204 205 // Inserts a new packet into NetEq. This is used by the InsertPacket method 206 // above. Returns 0 on success, otherwise an error code. 207 // TODO(hlundin): Merge this with InsertPacket above? 208 int InsertPacketInternal(const WebRtcRTPHeader& rtp_header, 209 rtc::ArrayView<const uint8_t> payload, 210 uint32_t receive_timestamp, 211 bool is_sync_packet) 212 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 213 214 // Delivers 10 ms of audio data. The data is written to |output|, which can 215 // hold (at least) |max_length| elements. The number of channels that were 216 // written to the output is provided in the output variable |num_channels|, 217 // and each channel contains |samples_per_channel| elements. If more than one 218 // channel is written, the samples are interleaved. 219 // Returns 0 on success, otherwise an error code. 220 int GetAudioInternal(size_t max_length, 221 int16_t* output, 222 size_t* samples_per_channel, 223 size_t* num_channels) 224 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 225 226 // Provides a decision to the GetAudioInternal method. The decision what to 227 // do is written to |operation|. Packets to decode are written to 228 // |packet_list|, and a DTMF event to play is written to |dtmf_event|. When 229 // DTMF should be played, |play_dtmf| is set to true by the method. 230 // Returns 0 on success, otherwise an error code. 231 int GetDecision(Operations* operation, 232 PacketList* packet_list, 233 DtmfEvent* dtmf_event, 234 bool* play_dtmf) EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 235 236 // Decodes the speech packets in |packet_list|, and writes the results to 237 // |decoded_buffer|, which is allocated to hold |decoded_buffer_length| 238 // elements. The length of the decoded data is written to |decoded_length|. 239 // The speech type -- speech or (codec-internal) comfort noise -- is written 240 // to |speech_type|. If |packet_list| contains any SID frames for RFC 3389 241 // comfort noise, those are not decoded. 242 int Decode(PacketList* packet_list, 243 Operations* operation, 244 int* decoded_length, 245 AudioDecoder::SpeechType* speech_type) 246 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 247 248 // Sub-method to Decode(). Performs codec internal CNG. 249 int DecodeCng(AudioDecoder* decoder, int* decoded_length, 250 AudioDecoder::SpeechType* speech_type) 251 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 252 253 // Sub-method to Decode(). Performs the actual decoding. 254 int DecodeLoop(PacketList* packet_list, 255 const Operations& operation, 256 AudioDecoder* decoder, 257 int* decoded_length, 258 AudioDecoder::SpeechType* speech_type) 259 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 260 261 // Sub-method which calls the Normal class to perform the normal operation. 262 void DoNormal(const int16_t* decoded_buffer, 263 size_t decoded_length, 264 AudioDecoder::SpeechType speech_type, 265 bool play_dtmf) EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 266 267 // Sub-method which calls the Merge class to perform the merge operation. 268 void DoMerge(int16_t* decoded_buffer, 269 size_t decoded_length, 270 AudioDecoder::SpeechType speech_type, 271 bool play_dtmf) EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 272 273 // Sub-method which calls the Expand class to perform the expand operation. 274 int DoExpand(bool play_dtmf) EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 275 276 // Sub-method which calls the Accelerate class to perform the accelerate 277 // operation. 278 int DoAccelerate(int16_t* decoded_buffer, 279 size_t decoded_length, 280 AudioDecoder::SpeechType speech_type, 281 bool play_dtmf, 282 bool fast_accelerate) EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 283 284 // Sub-method which calls the PreemptiveExpand class to perform the 285 // preemtive expand operation. 286 int DoPreemptiveExpand(int16_t* decoded_buffer, 287 size_t decoded_length, 288 AudioDecoder::SpeechType speech_type, 289 bool play_dtmf) EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 290 291 // Sub-method which calls the ComfortNoise class to generate RFC 3389 comfort 292 // noise. |packet_list| can either contain one SID frame to update the 293 // noise parameters, or no payload at all, in which case the previously 294 // received parameters are used. 295 int DoRfc3389Cng(PacketList* packet_list, bool play_dtmf) 296 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 297 298 // Calls the audio decoder to generate codec-internal comfort noise when 299 // no packet was received. 300 void DoCodecInternalCng(const int16_t* decoded_buffer, size_t decoded_length) 301 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 302 303 // Calls the DtmfToneGenerator class to generate DTMF tones. 304 int DoDtmf(const DtmfEvent& dtmf_event, bool* play_dtmf) 305 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 306 307 // Produces packet-loss concealment using alternative methods. If the codec 308 // has an internal PLC, it is called to generate samples. Otherwise, the 309 // method performs zero-stuffing. 310 void DoAlternativePlc(bool increase_timestamp) 311 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 312 313 // Overdub DTMF on top of |output|. 314 int DtmfOverdub(const DtmfEvent& dtmf_event, 315 size_t num_channels, 316 int16_t* output) const EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 317 318 // Extracts packets from |packet_buffer_| to produce at least 319 // |required_samples| samples. The packets are inserted into |packet_list|. 320 // Returns the number of samples that the packets in the list will produce, or 321 // -1 in case of an error. 322 int ExtractPackets(size_t required_samples, PacketList* packet_list) 323 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 324 325 // Resets various variables and objects to new values based on the sample rate 326 // |fs_hz| and |channels| number audio channels. 327 void SetSampleRateAndChannels(int fs_hz, size_t channels) 328 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 329 330 // Returns the output type for the audio produced by the latest call to 331 // GetAudio(). 332 NetEqOutputType LastOutputType() EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 333 334 // Updates Expand and Merge. 335 virtual void UpdatePlcComponents(int fs_hz, size_t channels) 336 EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 337 338 // Creates DecisionLogic object with the mode given by |playout_mode_|. 339 virtual void CreateDecisionLogic() EXCLUSIVE_LOCKS_REQUIRED(crit_sect_); 340 341 const rtc::scoped_ptr<CriticalSectionWrapper> crit_sect_; 342 const rtc::scoped_ptr<BufferLevelFilter> buffer_level_filter_ 343 GUARDED_BY(crit_sect_); 344 const rtc::scoped_ptr<DecoderDatabase> decoder_database_ 345 GUARDED_BY(crit_sect_); 346 const rtc::scoped_ptr<DelayManager> delay_manager_ GUARDED_BY(crit_sect_); 347 const rtc::scoped_ptr<DelayPeakDetector> delay_peak_detector_ 348 GUARDED_BY(crit_sect_); 349 const rtc::scoped_ptr<DtmfBuffer> dtmf_buffer_ GUARDED_BY(crit_sect_); 350 const rtc::scoped_ptr<DtmfToneGenerator> dtmf_tone_generator_ 351 GUARDED_BY(crit_sect_); 352 const rtc::scoped_ptr<PacketBuffer> packet_buffer_ GUARDED_BY(crit_sect_); 353 const rtc::scoped_ptr<PayloadSplitter> payload_splitter_ 354 GUARDED_BY(crit_sect_); 355 const rtc::scoped_ptr<TimestampScaler> timestamp_scaler_ 356 GUARDED_BY(crit_sect_); 357 const rtc::scoped_ptr<PostDecodeVad> vad_ GUARDED_BY(crit_sect_); 358 const rtc::scoped_ptr<ExpandFactory> expand_factory_ GUARDED_BY(crit_sect_); 359 const rtc::scoped_ptr<AccelerateFactory> accelerate_factory_ 360 GUARDED_BY(crit_sect_); 361 const rtc::scoped_ptr<PreemptiveExpandFactory> preemptive_expand_factory_ 362 GUARDED_BY(crit_sect_); 363 364 rtc::scoped_ptr<BackgroundNoise> background_noise_ GUARDED_BY(crit_sect_); 365 rtc::scoped_ptr<DecisionLogic> decision_logic_ GUARDED_BY(crit_sect_); 366 rtc::scoped_ptr<AudioMultiVector> algorithm_buffer_ GUARDED_BY(crit_sect_); 367 rtc::scoped_ptr<SyncBuffer> sync_buffer_ GUARDED_BY(crit_sect_); 368 rtc::scoped_ptr<Expand> expand_ GUARDED_BY(crit_sect_); 369 rtc::scoped_ptr<Normal> normal_ GUARDED_BY(crit_sect_); 370 rtc::scoped_ptr<Merge> merge_ GUARDED_BY(crit_sect_); 371 rtc::scoped_ptr<Accelerate> accelerate_ GUARDED_BY(crit_sect_); 372 rtc::scoped_ptr<PreemptiveExpand> preemptive_expand_ GUARDED_BY(crit_sect_); 373 RandomVector random_vector_ GUARDED_BY(crit_sect_); 374 rtc::scoped_ptr<ComfortNoise> comfort_noise_ GUARDED_BY(crit_sect_); 375 Rtcp rtcp_ GUARDED_BY(crit_sect_); 376 StatisticsCalculator stats_ GUARDED_BY(crit_sect_); 377 int fs_hz_ GUARDED_BY(crit_sect_); 378 int fs_mult_ GUARDED_BY(crit_sect_); 379 int last_output_sample_rate_hz_ GUARDED_BY(crit_sect_); 380 size_t output_size_samples_ GUARDED_BY(crit_sect_); 381 size_t decoder_frame_length_ GUARDED_BY(crit_sect_); 382 Modes last_mode_ GUARDED_BY(crit_sect_); 383 rtc::scoped_ptr<int16_t[]> mute_factor_array_ GUARDED_BY(crit_sect_); 384 size_t decoded_buffer_length_ GUARDED_BY(crit_sect_); 385 rtc::scoped_ptr<int16_t[]> decoded_buffer_ GUARDED_BY(crit_sect_); 386 uint32_t playout_timestamp_ GUARDED_BY(crit_sect_); 387 bool new_codec_ GUARDED_BY(crit_sect_); 388 uint32_t timestamp_ GUARDED_BY(crit_sect_); 389 bool reset_decoder_ GUARDED_BY(crit_sect_); 390 uint8_t current_rtp_payload_type_ GUARDED_BY(crit_sect_); 391 uint8_t current_cng_rtp_payload_type_ GUARDED_BY(crit_sect_); 392 uint32_t ssrc_ GUARDED_BY(crit_sect_); 393 bool first_packet_ GUARDED_BY(crit_sect_); 394 int error_code_ GUARDED_BY(crit_sect_); // Store last error code. 395 int decoder_error_code_ GUARDED_BY(crit_sect_); 396 const BackgroundNoiseMode background_noise_mode_ GUARDED_BY(crit_sect_); 397 NetEqPlayoutMode playout_mode_ GUARDED_BY(crit_sect_); 398 bool enable_fast_accelerate_ GUARDED_BY(crit_sect_); 399 rtc::scoped_ptr<Nack> nack_ GUARDED_BY(crit_sect_); 400 bool nack_enabled_ GUARDED_BY(crit_sect_); 401 402 private: 403 RTC_DISALLOW_COPY_AND_ASSIGN(NetEqImpl); 404 }; 405 406 } // namespace webrtc 407 #endif // WEBRTC_MODULES_AUDIO_CODING_NETEQ_NETEQ_IMPL_H_ 408