media/formats/mp4/aac.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/mp4/aac.h"
   6
   7 #include <algorithm>
   8
   9 #include "base/logging.h"
  10 #include "media/base/bit_reader.h"
  11 #include "media/base/media_log.h"
  12 #include "media/formats/mp4/rcheck.h"
  13 #include "media/formats/mpeg/adts_constants.h"
  14
  15 namespace media {
  16 namespace mp4 {
  17
  18 AAC::AAC()
  19     : profile_(0), frequency_index_(0), channel_config_(0), frequency_(0),
  20       extension_frequency_(0), channel_layout_(CHANNEL_LAYOUT_UNSUPPORTED) {
  21 }
  22
  23 AAC::~AAC() {
  24 }
  25
  26 bool AAC::Parse(const std::vector<uint8>& data, const LogCB& log_cb) {
  27 #if defined(OS_ANDROID)
  28   codec_specific_data_ = data;
  29 #endif
  30   if (data.empty())
  31     return false;
  32
  33   BitReader reader(&data[0], data.size());
  34   uint8 extension_type = 0;
  35   bool ps_present = false;
  36   uint8 extension_frequency_index = 0xff;
  37
  38   frequency_ = 0;
  39   extension_frequency_ = 0;
  40
  41   // The following code is written according to ISO 14496 Part 3 Table 1.13 -
  42   // Syntax of AudioSpecificConfig.
  43
  44   // Read base configuration
  45   RCHECK(reader.ReadBits(5, &profile_));
  46   RCHECK(reader.ReadBits(4, &frequency_index_));
  47   if (frequency_index_ == 0xf)
  48     RCHECK(reader.ReadBits(24, &frequency_));
  49   RCHECK(reader.ReadBits(4, &channel_config_));
  50
  51   // Read extension configuration.
  52   if (profile_ == 5 || profile_ == 29) {
  53     ps_present = (profile_ == 29);
  54     extension_type = 5;
  55     RCHECK(reader.ReadBits(4, &extension_frequency_index));
  56     if (extension_frequency_index == 0xf)
  57       RCHECK(reader.ReadBits(24, &extension_frequency_));
  58     RCHECK(reader.ReadBits(5, &profile_));
  59   }
  60
  61   MEDIA_LOG(log_cb) << "Audio codec: mp4a.40."
  62                     << std::hex << static_cast<int>(profile_);
  63
  64   RCHECK(SkipDecoderGASpecificConfig(&reader));
  65   RCHECK(SkipErrorSpecificConfig());
  66
  67   // Read extension configuration again
  68   // Note: The check for 16 available bits comes from the AAC spec.
  69   if (extension_type != 5 && reader.bits_available() >= 16) {
  70     uint16 sync_extension_type;
  71     uint8 sbr_present_flag;
  72     uint8 ps_present_flag;
  73
  74     if (reader.ReadBits(11, &sync_extension_type) &&
  75         sync_extension_type == 0x2b7) {
  76       if (reader.ReadBits(5, &extension_type) && extension_type == 5) {
  77         RCHECK(reader.ReadBits(1, &sbr_present_flag));
  78
  79         if (sbr_present_flag) {
  80           RCHECK(reader.ReadBits(4, &extension_frequency_index));
  81
  82           if (extension_frequency_index == 0xf)
  83             RCHECK(reader.ReadBits(24, &extension_frequency_));
  84
  85           // Note: The check for 12 available bits comes from the AAC spec.
  86           if (reader.bits_available() >= 12) {
  87             RCHECK(reader.ReadBits(11, &sync_extension_type));
  88             if (sync_extension_type == 0x548) {
  89               RCHECK(reader.ReadBits(1, &ps_present_flag));
  90               ps_present = ps_present_flag != 0;
  91             }
  92           }
  93         }
  94       }
  95     }
  96   }
  97
  98   if (frequency_ == 0) {
  99     RCHECK(frequency_index_ < kADTSFrequencyTableSize);
 100     frequency_ = kADTSFrequencyTable[frequency_index_];
 101   }
 102
 103   if (extension_frequency_ == 0 && extension_frequency_index != 0xff) {
 104     RCHECK(extension_frequency_index < kADTSFrequencyTableSize);
 105     extension_frequency_ = kADTSFrequencyTable[extension_frequency_index];
 106   }
 107
 108   // When Parametric Stereo is on, mono will be played as stereo.
 109   if (ps_present && channel_config_ == 1) {
 110     channel_layout_ = CHANNEL_LAYOUT_STEREO;
 111   } else {
 112     RCHECK(channel_config_ < kADTSChannelLayoutTableSize);
 113     channel_layout_ = kADTSChannelLayoutTable[channel_config_];
 114   }
 115
 116   return frequency_ != 0 && channel_layout_ != CHANNEL_LAYOUT_NONE &&
 117          profile_ >= 1 && profile_ <= 4;
 118 }
 119
 120 int AAC::GetOutputSamplesPerSecond(bool sbr_in_mimetype) const {
 121   if (extension_frequency_ > 0)
 122     return extension_frequency_;
 123
 124   if (!sbr_in_mimetype)
 125     return frequency_;
 126
 127   // The following code is written according to ISO 14496 Part 3 Table 1.11 and
 128   // Table 1.22. (Table 1.11 refers to the capping to 48000, Table 1.22 refers
 129   // to SBR doubling the AAC sample rate.)
 130   // TODO(acolwell) : Extend sample rate cap to 96kHz for Level 5 content.
 131   DCHECK_GT(frequency_, 0);
 132   return std::min(2 * frequency_, 48000);
 133 }
 134
 135 ChannelLayout AAC::GetChannelLayout(bool sbr_in_mimetype) const {
 136   // Check for implicit signalling of HE-AAC and indicate stereo output
 137   // if the mono channel configuration is signalled.
 138   // See ISO-14496-3 Section 1.6.6.1.2 for details about this special casing.
 139   if (sbr_in_mimetype && channel_config_ == 1)
 140     return CHANNEL_LAYOUT_STEREO;
 141
 142   return channel_layout_;
 143 }
 144
 145 bool AAC::ConvertEsdsToADTS(std::vector<uint8>* buffer) const {
 146   size_t size = buffer->size() + kADTSHeaderMinSize;
 147
 148   DCHECK(profile_ >= 1 && profile_ <= 4 && frequency_index_ != 0xf &&
 149          channel_config_ <= 7);
 150
 151   // ADTS header uses 13 bits for packet size.
 152   if (size >= (1 << 13))
 153     return false;
 154
 155   std::vector<uint8>& adts = *buffer;
 156
 157   adts.insert(buffer->begin(), kADTSHeaderMinSize, 0);
 158   adts[0] = 0xff;
 159   adts[1] = 0xf1;
 160   adts[2] = ((profile_ - 1) << 6) + (frequency_index_ << 2) +
 161       (channel_config_ >> 2);
 162   adts[3] = ((channel_config_ & 0x3) << 6) + (size >> 11);
 163   adts[4] = (size & 0x7ff) >> 3;
 164   adts[5] = ((size & 7) << 5) + 0x1f;
 165   adts[6] = 0xfc;
 166
 167   return true;
 168 }
 169
 170 // Currently this function only support GASpecificConfig defined in
 171 // ISO 14496 Part 3 Table 4.1 - Syntax of GASpecificConfig()
 172 bool AAC::SkipDecoderGASpecificConfig(BitReader* bit_reader) const {
 173   switch (profile_) {
 174     case 1:
 175     case 2:
 176     case 3:
 177     case 4:
 178     case 6:
 179     case 7:
 180     case 17:
 181     case 19:
 182     case 20:
 183     case 21:
 184     case 22:
 185     case 23:
 186       return SkipGASpecificConfig(bit_reader);
 187     default:
 188       break;
 189   }
 190
 191   return false;
 192 }
 193
 194 bool AAC::SkipErrorSpecificConfig() const {
 195   switch (profile_) {
 196     case 17:
 197     case 19:
 198     case 20:
 199     case 21:
 200     case 22:
 201     case 23:
 202     case 24:
 203     case 25:
 204     case 26:
 205     case 27:
 206       return false;
 207     default:
 208       break;
 209   }
 210
 211   return true;
 212 }
 213
 214 // The following code is written according to ISO 14496 part 3 Table 4.1 -
 215 // GASpecificConfig.
 216 bool AAC::SkipGASpecificConfig(BitReader* bit_reader) const {
 217   uint8 extension_flag = 0;
 218   uint8 depends_on_core_coder;
 219   uint16 dummy;
 220
 221   RCHECK(bit_reader->ReadBits(1, &dummy));  // frameLengthFlag
 222   RCHECK(bit_reader->ReadBits(1, &depends_on_core_coder));
 223   if (depends_on_core_coder == 1)
 224     RCHECK(bit_reader->ReadBits(14, &dummy));  // coreCoderDelay
 225
 226   RCHECK(bit_reader->ReadBits(1, &extension_flag));
 227   RCHECK(channel_config_ != 0);
 228
 229   if (profile_ == 6 || profile_ == 20)
 230     RCHECK(bit_reader->ReadBits(3, &dummy));  // layerNr
 231
 232   if (extension_flag) {
 233     if (profile_ == 22) {
 234       RCHECK(bit_reader->ReadBits(5, &dummy));  // numOfSubFrame
 235       RCHECK(bit_reader->ReadBits(11, &dummy));  // layer_length
 236     }
 237
 238     if (profile_ == 17 || profile_ == 19 || profile_ == 20 || profile_ == 23) {
 239       RCHECK(bit_reader->ReadBits(3, &dummy));  // resilience flags
 240     }
 241
 242     RCHECK(bit_reader->ReadBits(1, &dummy));  // extensionFlag3
 243   }
 244
 245   return true;
 246 }
 247
 248 }  // namespace mp4
 249
 250 }  // namespace media