Support DTS audio in MP4 - Part 1

Issue #52

Change-Id: I5061e2a674d5c1b039a49729e53d4d15a4120a5c

packager/media/base/audio_stream_info.cc

@@ -43,6 +43,18 @@ std::string AudioCodecToString(AudioCodec audio_codec) {
       return "Opus";
     case kCodecEAC3:
       return "EAC3";
+    case kCodecDTSC:
+      return "DTSC";
+    case kCodecDTSH:
+      return "DTSH";
+    case kCodecDTSL:
+      return "DTSL";
+    case kCodecDTSE:
+      return "DTSE";
+    case kCodecDTSP:
+      return "DTS+";
+    case kCodecDTSM:
+      return "DTS-";
     default:
       NOTIMPLEMENTED() << "Unknown Audio Codec: " << audio_codec;
       return "UnknownAudioCodec";
@@ -104,6 +116,18 @@ std::string AudioStreamInfo::GetCodecString(AudioCodec codec,
       return "opus";
     case kCodecAAC:
       return "mp4a.40." + base::UintToString(audio_object_type);
+    case kCodecDTSC:
+      return "dtsc";
+    case kCodecDTSH:
+      return "dtsh";
+    case kCodecDTSL:
+      return "dtsl";
+    case kCodecDTSE:
+      return "dtse";
+    case kCodecDTSP:
+      return "dts+";
+    case kCodecDTSM:
+      return "dts-";
     default:
       NOTIMPLEMENTED() << "Codec: " << codec;
       return "unknown";

packager/media/base/audio_stream_info.h

@@ -29,6 +29,12 @@ enum AudioCodec {
   kCodecPCM_S24BE,
   kCodecOpus,
   kCodecEAC3,
+  kCodecDTSC,
+  kCodecDTSH,
+  kCodecDTSL,
+  kCodecDTSE,
+  kCodecDTSP,
+  kCodecDTSM,
 
   kNumAudioCodec
 };

packager/media/formats/mp4/box_definitions.cc

@@ -1065,6 +1065,30 @@ uint32_t ElementaryStreamDescriptor::ComputeSize() {
   return atom_size;
 }
 
+DTSSpecificBox::DTSSpecificBox() {}
+DTSSpecificBox::~DTSSpecificBox() {}
+FourCC DTSSpecificBox::BoxType() const { return FOURCC_DDTS; }
+
+bool DTSSpecificBox::ReadWrite(BoxBuffer* buffer) {
+  RCHECK(Box::ReadWrite(buffer));
+
+  if (buffer->Reading()) {
+    RCHECK(
+        buffer->ReadWriteVector(&data, buffer->Size() - buffer->Pos()));
+  } else {
+    RCHECK(buffer->ReadWriteVector(&data, data.size()));
+  }
+  return true;
+}
+
+uint32_t DTSSpecificBox::ComputeSize() {
+  // This box is optional. Skip it if not initialized.
+  atom_size = 0;
+  if (data.size() > 0)
+    atom_size = kBoxSize + data.size();
+  return atom_size;
+}
+
 AudioSampleEntry::AudioSampleEntry()
     : format(FOURCC_NULL),
       data_reference_index(1),
@@ -1115,15 +1139,17 @@ bool AudioSampleEntry::ReadWrite(BoxBuffer* buffer) {
     }
   }
 
-  // ESDS is not valid in case of EAC3.
   RCHECK(buffer->TryReadWriteChild(&esds));
+  RCHECK(buffer->TryReadWriteChild(&ddts));
+
   return true;
 }
 
 uint32_t AudioSampleEntry::ComputeSize() {
   atom_size = kBoxSize + sizeof(data_reference_index) + sizeof(channelcount) +
               sizeof(samplesize) + sizeof(samplerate) + sinf.ComputeSize() +
-              esds.ComputeSize() + 6 + 8 +  // 6 + 8 bytes reserved.
+              esds.ComputeSize() + ddts.ComputeSize() +
+              6 + 8 +  // 6 + 8 bytes reserved.
               4;       // 4 bytes predefined.
   return atom_size;
 }

packager/media/formats/mp4/box_definitions.h

@@ -203,6 +203,12 @@ struct ElementaryStreamDescriptor : FullBox {
   ESDescriptor es_descriptor;
 };
 
+struct DTSSpecificBox : Box {
+  DECLARE_BOX_METHODS(DTSSpecificBox);
+
+  std::vector<uint8_t> data;
+};
+
 struct AudioSampleEntry : Box {
   DECLARE_BOX_METHODS(AudioSampleEntry);
   // Returns actual format of this sample entry.
@@ -218,6 +224,7 @@ struct AudioSampleEntry : Box {
 
   ProtectionSchemeInfo sinf;
   ElementaryStreamDescriptor esds;
+  DTSSpecificBox ddts;
 };
 
 struct SampleDescription : FullBox {

packager/media/formats/mp4/box_definitions_comparison.h

@@ -194,13 +194,19 @@ inline bool operator==(const ElementaryStreamDescriptor& lhs,
   return lhs.es_descriptor == rhs.es_descriptor;
 }
 
+inline bool operator==(const DTSSpecificBox& lhs,
+                       const DTSSpecificBox& rhs) {
+  return lhs.data == rhs.data;
+}
+
 inline bool operator==(const AudioSampleEntry& lhs,
                        const AudioSampleEntry& rhs) {
   return lhs.format == rhs.format &&
          lhs.data_reference_index == rhs.data_reference_index &&
          lhs.channelcount == rhs.channelcount &&
          lhs.samplesize == rhs.samplesize && lhs.samplerate == rhs.samplerate &&
-         lhs.sinf == rhs.sinf && lhs.esds == rhs.esds;
+         lhs.sinf == rhs.sinf && lhs.esds == rhs.esds &&
+         lhs.ddts == rhs.ddts;
 }
 
 inline bool operator==(const MediaHeader& lhs, const MediaHeader& rhs) {

packager/media/formats/mp4/box_definitions_unittest.cc

@@ -854,6 +854,25 @@ INSTANTIATE_TYPED_TEST_CASE_P(BoxDefinitionTypedTests2,
 // Test other cases of box input.
 class BoxDefinitionsTest : public BoxDefinitionsTestGeneral<Box> {};
 
+TEST_F(BoxDefinitionsTest, DTSSampleEntry) {
+  const uint8_t kDtseData[] = {0x00, 0x00, 0x00, 0x1c, 0x64, 0x64, 0x74,
+                               0x73, 0x00, 0x00, 0xbb, 0x80, 0x00, 0x03,
+                               0xe4, 0x18, 0x00, 0x03, 0xe4, 0x18, 0x18,
+                               0xe4, 0x7c, 0x00, 0x04, 0x00, 0x0f, 0x00};
+  AudioSampleEntry entry;
+  entry.format = FOURCC_DTSE;
+  entry.data_reference_index = 2;
+  entry.channelcount = 5;
+  entry.samplesize = 16;
+  entry.samplerate = 44100;
+  entry.ddts.data.assign(kDtseData, kDtseData + arraysize(kDtseData));
+  entry.Write(this->buffer_.get());
+
+  AudioSampleEntry entry_readback;
+  ASSERT_TRUE(ReadBack(&entry_readback));
+  ASSERT_EQ(entry, entry_readback);
+}
+
 TEST_F(BoxDefinitionsTest, ProtectionSystemSpecificHeader) {
   ProtectionSystemSpecificHeader pssh;
   Fill(&pssh);

packager/media/formats/mp4/es_descriptor.h

@@ -24,7 +24,6 @@ enum ObjectType {
   kForbidden = 0,
   kISO_14496_3 = 0x40,         // MPEG4 AAC
   kISO_13818_7_AAC_LC = 0x67,  // MPEG2 AAC-LC
-  kEAC3 = 0xa6                 // Dolby Digital Plus
 };
 
 /// This class parses object type and decoder specific information from an

packager/media/formats/mp4/fourccs.h

@@ -20,8 +20,15 @@ enum FourCC {
   FOURCC_CO64 = 0x636f3634,
   FOURCC_CTTS = 0x63747473,
   FOURCC_DASH = 0x64617368,
+  FOURCC_DDTS = 0x64647473,
   FOURCC_DINF = 0x64696e66,
   FOURCC_DREF = 0x64726566,
+  FOURCC_DTSC = 0x64747363,
+  FOURCC_DTSE = 0x64747365,
+  FOURCC_DTSH = 0x64747368,
+  FOURCC_DTSL = 0x6474736c,
+  FOURCC_DTSM = 0x6474732d,
+  FOURCC_DTSP = 0x6474732b,
   FOURCC_EAC3 = 0x65632d33,
   FOURCC_EDTS = 0x65647473,
   FOURCC_ELST = 0x656c7374,

packager/media/formats/mp4/mp4_media_parser.cc

@@ -40,7 +40,7 @@ uint64_t Rescale(uint64_t time_in_old_scale,
   return (static_cast<double>(time_in_old_scale) / old_scale) * new_scale;
 }
 
-VideoCodec FourCCToCodec(FourCC fourcc) {
+VideoCodec FourCCToVideoCodec(FourCC fourcc) {
   switch (fourcc) {
     case FOURCC_AVC1:
       return kCodecH264;
@@ -59,6 +59,27 @@ VideoCodec FourCCToCodec(FourCC fourcc) {
   }
 }
 
+AudioCodec FourCCToAudioCodec(FourCC fourcc) {
+  switch(fourcc) {
+    case FOURCC_DTSC:
+      return kCodecDTSC;
+    case FOURCC_DTSH:
+      return kCodecDTSH;
+    case FOURCC_DTSL:
+      return kCodecDTSL;
+    case FOURCC_DTSE:
+      return kCodecDTSE;
+    case FOURCC_DTSP:
+      return kCodecDTSP;
+    case FOURCC_DTSM:
+      return kCodecDTSM;
+    case FOURCC_EAC3:
+      return kCodecEAC3;
+    default:
+      return kUnknownAudioCodec;
+  }
+}
+
 const char kWidevineKeySystemId[] = "edef8ba979d64acea3c827dcd51d21ed";
 
 }  // namespace
@@ -310,27 +331,17 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
       // description indices, so we fail gracefully in that case.
       if (desc_idx >= samp_descr.audio_entries.size())
         desc_idx = 0;
+
       const AudioSampleEntry& entry = samp_descr.audio_entries[desc_idx];
-
+      const FourCC actual_format = entry.GetActualFormat();
-      if (!(entry.format == FOURCC_MP4A || entry.format == FOURCC_EAC3 ||
+      AudioCodec codec = FourCCToAudioCodec(actual_format);
-            (entry.format == FOURCC_ENCA &&
-             entry.sinf.format.format == FOURCC_MP4A))) {
-        LOG(ERROR) << "Unsupported audio format 0x"
-                   << std::hex << entry.format << " in stsd box.";
-        return false;
-      }
-
-      ObjectType audio_type = entry.esds.es_descriptor.object_type();
-      DVLOG(1) << "audio_type " << std::hex << audio_type;
-      if (audio_type == kForbidden && entry.format == FOURCC_EAC3) {
-        audio_type = kEAC3;
-      }
-
-      AudioCodec codec = kUnknownAudioCodec;
       uint8_t num_channels = 0;
       uint32_t sampling_frequency = 0;
       uint8_t audio_object_type = 0;
       std::vector<uint8_t> extra_data;
+
+      switch (actual_format) {
+        case FOURCC_MP4A:
           // Check if it is MPEG4 AAC defined in ISO 14496 Part 3 or
           // supported MPEG2 AAC variants.
           if (entry.esds.es_descriptor.IsAAC()) {
@@ -341,13 +352,28 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
             sampling_frequency = aac_audio_specific_config.frequency();
             audio_object_type = aac_audio_specific_config.audio_object_type();
             extra_data = entry.esds.es_descriptor.decoder_specific_info();
-      } else if (audio_type == kEAC3) {
+            break;
-        codec = kCodecEAC3;
+          } else {
+            LOG(ERROR) << "Unsupported audio format 0x" << std::hex
+                       << actual_format << " in stsd box.";
+            return false;
+          }
+        case FOURCC_DTSC:
+        case FOURCC_DTSH:
+        case FOURCC_DTSL:
+        case FOURCC_DTSE:
+        case FOURCC_DTSM:
+          extra_data = entry.ddts.data;
           num_channels = entry.channelcount;
           sampling_frequency = entry.samplerate;
-      } else {
+          break;
-        LOG(ERROR) << "Unsupported audio object type 0x"
+        case FOURCC_EAC3:
-                   << std::hex << audio_type << " in esds.";
+          num_channels = entry.channelcount;
+          sampling_frequency = entry.samplerate;
+          break;
+        default:
+          LOG(ERROR) << "Unsupported audio format 0x" << std::hex
+                     << actual_format << " in stsd box.";
           return false;
       }
 
@@ -386,7 +412,7 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
       uint8_t nalu_length_size = 0;
 
       const FourCC actual_format = entry.GetActualFormat();
-      const VideoCodec video_codec = FourCCToCodec(actual_format);
+      const VideoCodec video_codec = FourCCToVideoCodec(actual_format);
       switch (actual_format) {
         case FOURCC_AVC1: {
           AVCDecoderConfiguration avc_config;