media/formats/mp2t/es_parser_adts.cc

   1 // Copyright 2014 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/formats/mp2t/es_parser_adts.h"
   6
   7 #include "base/basictypes.h"
   8 #include "base/logging.h"
   9 #include "base/strings/string_number_conversions.h"
  10 #include "media/base/audio_timestamp_helper.h"
  11 #include "media/base/bit_reader.h"
  12 #include "media/base/channel_layout.h"
  13 #include "media/base/stream_parser_buffer.h"
  14 #include "media/base/timestamp_constants.h"
  15 #include "media/formats/common/offset_byte_queue.h"
  16 #include "media/formats/mp2t/mp2t_common.h"
  17 #include "media/formats/mpeg/adts_constants.h"
  18
  19 namespace media {
  20
  21 static int ExtractAdtsFrameSize(const uint8* adts_header) {
  22   return ((static_cast<int>(adts_header[5]) >> 5) |
  23           (static_cast<int>(adts_header[4]) << 3) |
  24           ((static_cast<int>(adts_header[3]) & 0x3) << 11));
  25 }
  26
  27 static size_t ExtractAdtsFrequencyIndex(const uint8* adts_header) {
  28   return ((adts_header[2] >> 2) & 0xf);
  29 }
  30
  31 static size_t ExtractAdtsChannelConfig(const uint8* adts_header) {
  32   return (((adts_header[3] >> 6) & 0x3) |
  33           ((adts_header[2] & 0x1) << 2));
  34 }
  35
  36 // Return true if buf corresponds to an ADTS syncword.
  37 // |buf| size must be at least 2.
  38 static bool isAdtsSyncWord(const uint8* buf) {
  39   // The first 12 bits must be 1.
  40   // The layer field (2 bits) must be set to 0.
  41   return (buf[0] == 0xff) && ((buf[1] & 0xf6) == 0xf0);
  42 }
  43
  44 namespace mp2t {
  45
  46 struct EsParserAdts::AdtsFrame {
  47   // Pointer to the ES data.
  48   const uint8* data;
  49
  50   // Frame size;
  51   int size;
  52
  53   // Frame offset in the ES queue.
  54   int64 queue_offset;
  55 };
  56
  57 bool EsParserAdts::LookForAdtsFrame(AdtsFrame* adts_frame) {
  58   int es_size;
  59   const uint8* es;
  60   es_queue_->Peek(&es, &es_size);
  61
  62   int max_offset = es_size - kADTSHeaderMinSize;
  63   if (max_offset <= 0)
  64     return false;
  65
  66   for (int offset = 0; offset < max_offset; offset++) {
  67     const uint8* cur_buf = &es[offset];
  68     if (!isAdtsSyncWord(cur_buf))
  69       continue;
  70
  71     int frame_size = ExtractAdtsFrameSize(cur_buf);
  72     if (frame_size < kADTSHeaderMinSize) {
  73       // Too short to be an ADTS frame.
  74       continue;
  75     }
  76
  77     int remaining_size = es_size - offset;
  78     if (remaining_size < frame_size) {
  79       // Not a full frame: will resume when we have more data.
  80       es_queue_->Pop(offset);
  81       return false;
  82     }
  83
  84     // Check whether there is another frame
  85     // |size| apart from the current one.
  86     if (remaining_size >= frame_size + 2 &&
  87         !isAdtsSyncWord(&cur_buf[frame_size])) {
  88       continue;
  89     }
  90
  91     es_queue_->Pop(offset);
  92     es_queue_->Peek(&adts_frame->data, &es_size);
  93     adts_frame->queue_offset = es_queue_->head();
  94     adts_frame->size = frame_size;
  95     DVLOG(LOG_LEVEL_ES)
  96         << "ADTS syncword @ pos=" << adts_frame->queue_offset
  97         << " frame_size=" << adts_frame->size;
  98     DVLOG(LOG_LEVEL_ES)
  99         << "ADTS header: "
 100         << base::HexEncode(adts_frame->data, kADTSHeaderMinSize);
 101     return true;
 102   }
 103
 104   es_queue_->Pop(max_offset);
 105   return false;
 106 }
 107
 108 void EsParserAdts::SkipAdtsFrame(const AdtsFrame& adts_frame) {
 109   DCHECK_EQ(adts_frame.queue_offset, es_queue_->head());
 110   es_queue_->Pop(adts_frame.size);
 111 }
 112
 113 EsParserAdts::EsParserAdts(
 114     const NewAudioConfigCB& new_audio_config_cb,
 115     const EmitBufferCB& emit_buffer_cb,
 116     bool sbr_in_mimetype)
 117   : new_audio_config_cb_(new_audio_config_cb),
 118     emit_buffer_cb_(emit_buffer_cb),
 119     sbr_in_mimetype_(sbr_in_mimetype) {
 120 }
 121
 122 EsParserAdts::~EsParserAdts() {
 123 }
 124
 125 bool EsParserAdts::ParseFromEsQueue() {
 126   // Look for every ADTS frame in the ES buffer.
 127   AdtsFrame adts_frame;
 128   while (LookForAdtsFrame(&adts_frame)) {
 129     // Update the audio configuration if needed.
 130     DCHECK_GE(adts_frame.size, kADTSHeaderMinSize);
 131     if (!UpdateAudioConfiguration(adts_frame.data))
 132       return false;
 133
 134     // Get the PTS & the duration of this access unit.
 135     TimingDesc current_timing_desc =
 136         GetTimingDescriptor(adts_frame.queue_offset);
 137     if (current_timing_desc.pts != kNoTimestamp())
 138       audio_timestamp_helper_->SetBaseTimestamp(current_timing_desc.pts);
 139
 140     if (audio_timestamp_helper_->base_timestamp() == kNoTimestamp()) {
 141       DVLOG(1) << "Skipping audio frame with unknown timestamp";
 142       SkipAdtsFrame(adts_frame);
 143       continue;
 144     }
 145     base::TimeDelta current_pts = audio_timestamp_helper_->GetTimestamp();
 146     base::TimeDelta frame_duration =
 147         audio_timestamp_helper_->GetFrameDuration(kSamplesPerAACFrame);
 148
 149     // Emit an audio frame.
 150     bool is_key_frame = true;
 151
 152     // TODO(wolenetz/acolwell): Validate and use a common cross-parser TrackId
 153     // type and allow multiple audio tracks. See https://crbug.com/341581.
 154     scoped_refptr<StreamParserBuffer> stream_parser_buffer =
 155         StreamParserBuffer::CopyFrom(
 156             adts_frame.data,
 157             adts_frame.size,
 158             is_key_frame,
 159             DemuxerStream::AUDIO, 0);
 160     stream_parser_buffer->set_timestamp(current_pts);
 161     stream_parser_buffer->SetDecodeTimestamp(
 162         DecodeTimestamp::FromPresentationTime(current_pts));
 163     stream_parser_buffer->set_duration(frame_duration);
 164     emit_buffer_cb_.Run(stream_parser_buffer);
 165
 166     // Update the PTS of the next frame.
 167     audio_timestamp_helper_->AddFrames(kSamplesPerAACFrame);
 168
 169     // Skip the current frame.
 170     SkipAdtsFrame(adts_frame);
 171   }
 172
 173   return true;
 174 }
 175
 176 void EsParserAdts::Flush() {
 177 }
 178
 179 void EsParserAdts::ResetInternal() {
 180   last_audio_decoder_config_ = AudioDecoderConfig();
 181 }
 182
 183 bool EsParserAdts::UpdateAudioConfiguration(const uint8* adts_header) {
 184   size_t frequency_index = ExtractAdtsFrequencyIndex(adts_header);
 185   if (frequency_index >= kADTSFrequencyTableSize) {
 186     // Frequency index 13 & 14 are reserved
 187     // while 15 means that the frequency is explicitly written
 188     // (not supported).
 189     return false;
 190   }
 191
 192   size_t channel_configuration = ExtractAdtsChannelConfig(adts_header);
 193   if (channel_configuration == 0 ||
 194       channel_configuration >= kADTSChannelLayoutTableSize) {
 195     // TODO(damienv): Add support for inband channel configuration.
 196     return false;
 197   }
 198
 199   // TODO(damienv): support HE-AAC frequency doubling (SBR)
 200   // based on the incoming ADTS profile.
 201   int samples_per_second = kADTSFrequencyTable[frequency_index];
 202   int adts_profile = (adts_header[2] >> 6) & 0x3;
 203
 204   // The following code is written according to ISO 14496 Part 3 Table 1.11 and
 205   // Table 1.22. (Table 1.11 refers to the capping to 48000, Table 1.22 refers
 206   // to SBR doubling the AAC sample rate.)
 207   // TODO(damienv) : Extend sample rate cap to 96kHz for Level 5 content.
 208   int extended_samples_per_second = sbr_in_mimetype_
 209       ? std::min(2 * samples_per_second, 48000)
 210       : samples_per_second;
 211
 212   // The following code is written according to ISO 14496 Part 3 Table 1.13 -
 213   // Syntax of AudioSpecificConfig.
 214   uint16 extra_data_int = static_cast<uint16>(
 215       // Note: adts_profile is in the range [0,3], since the ADTS header only
 216       // allows two bits for its value.
 217       ((adts_profile + 1) << 11) +
 218       // frequency_index is [0..13], per early out above.
 219       (frequency_index << 7) +
 220       // channel_configuration is [0..7], per early out above.
 221       (channel_configuration << 3));
 222   uint8 extra_data[2] = {
 223       static_cast<uint8>(extra_data_int >> 8),
 224       static_cast<uint8>(extra_data_int & 0xff)
 225   };
 226
 227   AudioDecoderConfig audio_decoder_config(
 228       kCodecAAC,
 229       kSampleFormatS16,
 230       kADTSChannelLayoutTable[channel_configuration],
 231       extended_samples_per_second,
 232       extra_data,
 233       arraysize(extra_data),
 234       false);
 235
 236   if (!audio_decoder_config.Matches(last_audio_decoder_config_)) {
 237     DVLOG(1) << "Sampling frequency: " << samples_per_second;
 238     DVLOG(1) << "Extended sampling frequency: " << extended_samples_per_second;
 239     DVLOG(1) << "Channel config: " << channel_configuration;
 240     DVLOG(1) << "Adts profile: " << adts_profile;
 241     // Reset the timestamp helper to use a new time scale.
 242     if (audio_timestamp_helper_ &&
 243         audio_timestamp_helper_->base_timestamp() != kNoTimestamp()) {
 244       base::TimeDelta base_timestamp = audio_timestamp_helper_->GetTimestamp();
 245       audio_timestamp_helper_.reset(
 246         new AudioTimestampHelper(samples_per_second));
 247       audio_timestamp_helper_->SetBaseTimestamp(base_timestamp);
 248     } else {
 249       audio_timestamp_helper_.reset(
 250           new AudioTimestampHelper(samples_per_second));
 251     }
 252     // Audio config notification.
 253     last_audio_decoder_config_ = audio_decoder_config;
 254     new_audio_config_cb_.Run(audio_decoder_config);
 255   }
 256
 257   return true;
 258 }
 259
 260 }  // namespace mp2t
 261 }  // namespace media