OLD | NEW |
(Empty) | |
| 1 // Copyright (c) 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 "media/mpeg2/es_parser_adts.h" |
| 6 |
| 7 #include <list> |
| 8 |
| 9 #include "base/basictypes.h" |
| 10 #include "base/logging.h" |
| 11 #include "base/strings/string_number_conversions.h" |
| 12 #include "media/base/audio_decoder_config.h" |
| 13 #include "media/base/bit_reader.h" |
| 14 #include "media/base/channel_layout.h" |
| 15 #include "media/base/stream_parser_buffer.h" |
| 16 #include "media/mpeg2/mpeg2ts_common.h" |
| 17 |
| 18 namespace { |
| 19 // Adts header is at least 7 bytes (can be 9 bytes). |
| 20 const int kAdtsHeaderMinSize = 7; |
| 21 |
| 22 const int adts_frequency_table[16] = { |
| 23 96000, |
| 24 88200, |
| 25 64000, |
| 26 48000, |
| 27 44100, |
| 28 32000, |
| 29 24000, |
| 30 22050, |
| 31 16000, |
| 32 12000, |
| 33 11025, |
| 34 8000, |
| 35 7350, |
| 36 0, |
| 37 0, |
| 38 0, |
| 39 }; |
| 40 const int kExplicitFrequencyIndex = 15; |
| 41 |
| 42 media::ChannelLayout adts_channel_layout[8] = { |
| 43 media::CHANNEL_LAYOUT_NONE, |
| 44 media::CHANNEL_LAYOUT_MONO, |
| 45 media::CHANNEL_LAYOUT_STEREO, |
| 46 media::CHANNEL_LAYOUT_SURROUND, |
| 47 media::CHANNEL_LAYOUT_4_0, |
| 48 media::CHANNEL_LAYOUT_5_0_BACK, |
| 49 media::CHANNEL_LAYOUT_5_1_BACK, |
| 50 media::CHANNEL_LAYOUT_7_1, |
| 51 }; |
| 52 |
| 53 // Number of samples per frame. |
| 54 const int kNumberSamplesPerAACFrame = 1024; |
| 55 const int kNumberSamplesPerHeAACFrame = 2048; |
| 56 const int kNumberSamplesPerAACLcFrame = 960; |
| 57 |
| 58 int ExtractAdtsFrameSize(const uint8* adts_header) { |
| 59 int frame_size = |
| 60 (static_cast<int>(adts_header[5]) >> 5) | |
| 61 (static_cast<int>(adts_header[4]) << 3) | |
| 62 ((static_cast<int>(adts_header[3]) & 0x3) << 11); |
| 63 return frame_size; |
| 64 } |
| 65 |
| 66 int ExtractAdtsFrequencyIndex(const uint8* adts_header) { |
| 67 int frequency_index = |
| 68 (adts_header[2] >> 2) & 0xf; |
| 69 return frequency_index; |
| 70 } |
| 71 |
| 72 int ExtractAdtsChannelConfig(const uint8* adts_header) { |
| 73 int channel_config = |
| 74 ((adts_header[3] >> 6) & 0x3) | |
| 75 ((adts_header[2] & 0x1) << 2); |
| 76 return channel_config; |
| 77 } |
| 78 |
| 79 // Look for an ADTS syncword. |
| 80 bool LookForSyncWord(const std::vector<uint8>& buf, |
| 81 int pos, |
| 82 int* new_pos, int* frame_sz) { |
| 83 int max_offset = buf.size() - kAdtsHeaderMinSize; |
| 84 if (max_offset < 0) { |
| 85 max_offset = 0; |
| 86 } |
| 87 |
| 88 for (int offset = pos; offset < max_offset; offset++) { |
| 89 const uint8* cur_buf = &buf[offset]; |
| 90 |
| 91 if ((cur_buf[0] != 0xff) || ((cur_buf[1] & 0xf6) != 0xf0)) { |
| 92 // The first 12 bits must be 1. |
| 93 // The layer field (2 bits) must be set to 0. |
| 94 continue; |
| 95 } |
| 96 |
| 97 int frequency_index = ExtractAdtsFrequencyIndex(cur_buf); |
| 98 if (frequency_index == kExplicitFrequencyIndex) { |
| 99 // 15 is a forbidden value. |
| 100 continue; |
| 101 } |
| 102 |
| 103 int frame_size = ExtractAdtsFrameSize(cur_buf); |
| 104 if (frame_size < kAdtsHeaderMinSize) { |
| 105 // Too short to be an ADTS frame. |
| 106 continue; |
| 107 } |
| 108 |
| 109 // Check whether there is another frame |
| 110 // |size| apart from the current one. |
| 111 int remaining_size = buf.size() - offset; |
| 112 if (remaining_size >= frame_size + 2) { |
| 113 if ((cur_buf[frame_size] != 0xff) || |
| 114 (cur_buf[frame_size + 1] & 0xf6) != 0xf0) { |
| 115 continue; |
| 116 } |
| 117 } |
| 118 |
| 119 *new_pos = offset; |
| 120 *frame_sz = frame_size; |
| 121 return true; |
| 122 } |
| 123 |
| 124 *new_pos = max_offset; |
| 125 return false; |
| 126 } |
| 127 |
| 128 } // namespace |
| 129 |
| 130 namespace media { |
| 131 namespace mpeg2ts { |
| 132 |
| 133 EsParserAdts::EsParserAdts( |
| 134 NewAudioConfigCB new_audio_config_cb, |
| 135 EmitBufferCB emit_buffer_cb) |
| 136 : first_frame_(true), |
| 137 new_audio_config_cb_(new_audio_config_cb), |
| 138 emit_buffer_cb_(emit_buffer_cb), |
| 139 is_audio_config_known_(false), |
| 140 sampling_frequency_(0), |
| 141 channel_configuration_(0) { |
| 142 } |
| 143 |
| 144 EsParserAdts::~EsParserAdts() { |
| 145 } |
| 146 |
| 147 void EsParserAdts::Parse(const uint8* buf, int size, |
| 148 bool is_pts_valid, base::TimeDelta pts, |
| 149 bool is_dts_valid, base::TimeDelta dts) { |
| 150 // The incoming PTS applies to the access unit that comes just after |
| 151 // the beginning of |buf|. |
| 152 if (is_pts_valid) { |
| 153 pts_list_.push_back(EsPts(raw_es_.size(), pts)); |
| 154 } |
| 155 |
| 156 // Copy the input data to the ES buffer. |
| 157 int old_size = raw_es_.size(); |
| 158 raw_es_.resize(old_size + size); |
| 159 memcpy(&raw_es_[old_size], buf, size); |
| 160 |
| 161 // Look for every ADTS frame in the ES buffer starting at offset = 0 |
| 162 int es_position = 0; |
| 163 int frame_size; |
| 164 while (LookForSyncWord(raw_es_, es_position, |
| 165 &es_position, &frame_size)) { |
| 166 VLOG(LOG_LEVEL_ES) << "ADTS syncword @ pos=" << es_position |
| 167 << " frame_size=" << frame_size; |
| 168 VLOG(LOG_LEVEL_ES) << "ADTS header: " |
| 169 << base::HexEncode(&raw_es_[es_position], 7); |
| 170 |
| 171 // Do not process the frame if this one is a partial frame. |
| 172 int remaining_size = raw_es_.size() - es_position; |
| 173 if (frame_size > remaining_size) { |
| 174 break; |
| 175 } |
| 176 |
| 177 // Update the audio configuration if needed. |
| 178 DCHECK_GE(frame_size, kAdtsHeaderMinSize); |
| 179 UpdateAudioConfiguration(&raw_es_[es_position]); |
| 180 |
| 181 // Get the PTS of this access unit. |
| 182 base::TimeDelta current_pts = estimated_pts_; |
| 183 while (!pts_list_.empty() && |
| 184 pts_list_.front().first <= es_position) { |
| 185 current_pts = pts_list_.front().second; |
| 186 pts_list_.pop_front(); |
| 187 } |
| 188 VLOG(LOG_LEVEL_ES) |
| 189 << "Current PTS: " << current_pts.InMilliseconds() |
| 190 << " Estimated PTS: " << estimated_pts_.InMilliseconds(); |
| 191 |
| 192 // Verify that PTS is increasing. |
| 193 if (!first_frame_ && current_pts < last_frame_pts_) { |
| 194 LOG(WARNING) << "ADTS: pts not monotonic"; |
| 195 } |
| 196 first_frame_ = false; |
| 197 last_frame_pts_ = current_pts; |
| 198 |
| 199 // Emit an audio frame. |
| 200 bool is_key_frame = true; |
| 201 scoped_refptr<StreamParserBuffer> stream_parser_buffer = |
| 202 StreamParserBuffer::CopyFrom( |
| 203 &raw_es_[es_position], |
| 204 frame_size, |
| 205 is_key_frame); |
| 206 stream_parser_buffer->SetDecodeTimestamp(current_pts); |
| 207 stream_parser_buffer->set_timestamp(current_pts); |
| 208 emit_buffer_cb_.Run(stream_parser_buffer); |
| 209 |
| 210 // Update the PTS of the next frame. |
| 211 base::TimeDelta frame_duration = |
| 212 base::TimeDelta::FromMicroseconds( |
| 213 (1000000 * kNumberSamplesPerAACFrame) / sampling_frequency_); |
| 214 estimated_pts_ = current_pts + frame_duration; |
| 215 |
| 216 // Skip the current frame. |
| 217 es_position += frame_size; |
| 218 } |
| 219 |
| 220 // Discard all the bytes that have been processed. |
| 221 DiscardEs(es_position); |
| 222 } |
| 223 |
| 224 void EsParserAdts::Flush() { |
| 225 // All the complete frames have been emitted, |
| 226 // so just clear the ES buffer. |
| 227 raw_es_.clear(); |
| 228 pts_list_.clear(); |
| 229 } |
| 230 |
| 231 void EsParserAdts::UpdateAudioConfiguration(const uint8* adts_header) { |
| 232 int frequency_index = ExtractAdtsFrequencyIndex(adts_header); |
| 233 if (frequency_index > 12) { |
| 234 // Frequency index 13 & 14 are reserved |
| 235 // while 15 means that the frequency is explicitly written |
| 236 // (not supported). |
| 237 return; |
| 238 } |
| 239 int samples_per_second = adts_frequency_table[frequency_index]; |
| 240 |
| 241 int channel_configuration = ExtractAdtsChannelConfig(adts_header); |
| 242 int adts_profile = (adts_header[2] >> 6) & 0x3; |
| 243 |
| 244 #if 0 |
| 245 // TODO(damienv): support HE-AAC frequency doubling (SBR) |
| 246 if (adts_profile == kAdtsProfileHeAAC) { |
| 247 samples_per_second *= 2; |
| 248 } |
| 249 #endif |
| 250 |
| 251 if (!is_audio_config_known_ || |
| 252 sampling_frequency_ != samples_per_second || |
| 253 channel_configuration_ != channel_configuration) { |
| 254 is_audio_config_known_ = true; |
| 255 sampling_frequency_ = samples_per_second; |
| 256 channel_configuration_ = channel_configuration; |
| 257 |
| 258 LOG(INFO) << "Sampling frequency: " << samples_per_second; |
| 259 LOG(INFO) << "Channel config: " << channel_configuration; |
| 260 LOG(INFO) << "Adts profile: " << adts_profile; |
| 261 AudioDecoderConfig audio_decoder_config( |
| 262 kCodecAAC, |
| 263 kSampleFormatS16, |
| 264 adts_channel_layout[channel_configuration], |
| 265 samples_per_second, |
| 266 NULL, 0, |
| 267 false); |
| 268 new_audio_config_cb_.Run(audio_decoder_config); |
| 269 } |
| 270 } |
| 271 |
| 272 void EsParserAdts::DiscardEs(int nbytes) { |
| 273 if (nbytes <= 0) { |
| 274 return; |
| 275 } |
| 276 |
| 277 // Adjust the ES position of each PTS. |
| 278 EsPtsList::iterator it = pts_list_.begin(); |
| 279 for (; it != pts_list_.end(); ++it) { |
| 280 it->first -= nbytes; |
| 281 } |
| 282 |
| 283 // Discard |nbytes| of ES. |
| 284 int old_size = raw_es_.size(); |
| 285 int new_size = old_size - nbytes; |
| 286 CHECK_LE(nbytes, old_size); |
| 287 if (new_size > 0) { |
| 288 memmove(&raw_es_[0], &raw_es_[nbytes], new_size); |
| 289 } |
| 290 raw_es_.resize(new_size); |
| 291 } |
| 292 |
| 293 } // namespace mpeg2ts |
| 294 } // namespace media |
OLD | NEW |