Support parsing MPEG-H 3D Audio

First intention to support parsing MPEG-H 3D Audio in ExoPlayer is to take advantage of parsing capability from MediaParser API in AOSP. Just with this change ExoPlayer does't support decoding MPEG-H 3D Audio but can support decoding either by adding decoder with an extension or by using Android OS which has decoder capability with MediaCodec API.
google · May 10, 2021 · f05197c · f05197c
1 parent 7f2922d
commit f05197c
Show file tree

Hide file tree

Showing 16 changed files with 1,994 additions and 1 deletion.
diff --git a/library/common/src/main/java/com/google/android/exoplayer2/util/MimeTypes.java b/library/common/src/main/java/com/google/android/exoplayer2/util/MimeTypes.java
@@ -83,6 +83,8 @@ public final class MimeTypes {
  public static final String AUDIO_MSGSM = BASE_TYPE_AUDIO + "/gsm";
  public static final String AUDIO_OGG = BASE_TYPE_AUDIO + "/ogg";
  public static final String AUDIO_WAV = BASE_TYPE_AUDIO + "/wav";
+ public static final String AUDIO_MPEGH_MHA1 = BASE_TYPE_AUDIO + "/mha1";
+ public static final String AUDIO_MPEGH_MHM1 = BASE_TYPE_AUDIO + "/mhm1";
  public static final String AUDIO_UNKNOWN = BASE_TYPE_AUDIO + "/x-unknown";
 
  public static final String TEXT_VTT = BASE_TYPE_TEXT + "/vtt";
@@ -391,6 +393,10 @@ public static String getMediaMimeType(@Nullable String codec) {
  return MimeTypes.APPLICATION_CEA708;
  } else if (codec.contains("eia608") || codec.contains("cea608")) {
  return MimeTypes.APPLICATION_CEA608;
+ } else if (codec.startsWith("mha1")) {
+ return MimeTypes.AUDIO_MPEGH_MHA1;
+ } else if (codec.startsWith("mhm1")) {
+ return MimeTypes.AUDIO_MPEGH_MHM1;
  } else {
  return getCustomMimeTypeForCodec(codec);
  }

diff --git a/library/extractor/src/main/java/com/google/android/exoplayer2/extractor/mp4/Atom.java b/library/extractor/src/main/java/com/google/android/exoplayer2/extractor/mp4/Atom.java
@@ -387,6 +387,15 @@
  @SuppressWarnings("ConstantCaseForConstants")
  public static final int TYPE_twos = 0x74776f73;
 
+ @SuppressWarnings("ConstantCaseForConstants")
+ public static final int TYPE_mha1 = 0x6d686131;
+
+ @SuppressWarnings("ConstantCaseForConstants")
+ public static final int TYPE_mhm1 = 0x6d686d31;
+
+ @SuppressWarnings("ConstantCaseForConstants")
+ public static final int TYPE_mhaC = 0x6d686143;
+
  public final int type;
 
  public Atom(int type) {

diff --git a/library/extractor/src/main/java/com/google/android/exoplayer2/extractor/mp4/AtomParsers.java b/library/extractor/src/main/java/com/google/android/exoplayer2/extractor/mp4/AtomParsers.java
@@ -966,7 +966,9 @@ private static StsdData parseStsd(
  || childAtomType == Atom.TYPE_alaw
  || childAtomType == Atom.TYPE_ulaw
  || childAtomType == Atom.TYPE_Opus
- || childAtomType == Atom.TYPE_fLaC) {
+ || childAtomType == Atom.TYPE_fLaC
+ || childAtomType == Atom.TYPE_mha1
+ || childAtomType == Atom.TYPE_mhm1) {
  parseAudioSampleEntry(
  stsd,
  childAtomType,
@@ -1358,6 +1360,10 @@ private static void parseAudioSampleEntry(
  mimeType = MimeTypes.AUDIO_OPUS;
  } else if (atomType == Atom.TYPE_fLaC) {
  mimeType = MimeTypes.AUDIO_FLAC;
+ } else if (atomType == Atom.TYPE_mha1) {
+ mimeType = MimeTypes.AUDIO_MPEGH_MHA1;
+ } else if (atomType == Atom.TYPE_mhm1) {
+ mimeType = MimeTypes.AUDIO_MPEGH_MHM1;
  }
 
  @Nullable List<byte[]> initializationData = null;
@@ -1440,6 +1446,16 @@ private static void parseAudioSampleEntry(
  sampleRate = audioSpecificConfig.first;
  channelCount = audioSpecificConfig.second;
  initializationData = ImmutableList.of(initializationDataBytes);
+ } else if (childAtomType == Atom.TYPE_mhaC) {
+ // See ISO_IEC_23008-3;2019 MHADecoderConfigurationRecord
+ int mhacHeaderSize = 4 /* size */ + 4 /* boxtype 'mhaC' */
+ + 1 /* configurationVersion */ + 1 /* mpegh3daProfileLevelIndication */
+ + 1 /* referenceChannelLayout */ + 2 /* mpegh3daConfigLength */;
+ int childAtomBodySize = childAtomSize - mhacHeaderSize;
+ byte[] initializationDataBytes = new byte[childAtomBodySize];
+ parent.setPosition(childPosition + mhacHeaderSize);
+ parent.readBytes(initializationDataBytes, 0, childAtomBodySize);
+ initializationData = ImmutableList.of(initializationDataBytes);
  }
  childPosition += childAtomSize;
  }

diff --git a/...extractor/src/test/java/com/google/android/exoplayer2/extractor/mp4/Mp4ExtractorTest.java b/...extractor/src/test/java/com/google/android/exoplayer2/extractor/mp4/Mp4ExtractorTest.java
@@ -84,4 +84,16 @@ public void mp4SampleWithOpusTrack() throws Exception {
  ExtractorAsserts.assertBehavior(
  Mp4Extractor::new, "media/mp4/sample_opus.mp4", simulationConfig);
  }
+
+ @Test
+ public void mp4SampleWithMha1Track() throws Exception {
+ ExtractorAsserts.assertBehavior(
+ Mp4Extractor::new, "media/mp4/sample_mpegh_mha1.mp4", simulationConfig);
+ }
+
+ @Test
+ public void mp4SampleWithMhm1Track() throws Exception {
+ ExtractorAsserts.assertBehavior(
+ Mp4Extractor::new, "media/mp4/sample_mpegh_mhm1.mp4", simulationConfig);
+ }
 }
diff --git a/testdata/src/test/assets/extractordumps/mp4/sample_mpegh_mha1.mp4.0.dump b/testdata/src/test/assets/extractordumps/mp4/sample_mpegh_mha1.mp4.0.dump
@@ -0,0 +1,255 @@
+seekMap:
+ isSeekable = true
+ duration = 1168020
+ getPosition(0) = [[timeUs=0, position=1256]]
+ getPosition(1) = [[timeUs=1, position=1256]]
+ getPosition(584010) = [[timeUs=584010, position=46570]]
+ getPosition(1168020) = [[timeUs=1168020, position=91172]]
+numberOfTracks = 1
+track 0:
+ total output bytes = 104853
+ sample count = 58
+ format 0:
+ id = 1
+ sampleMimeType = audio/mha1
+ maxInputSize = 3528
+ channelCount = 0
+ sampleRate = 48000
+ encoderDelay = 3072
+ encoderPadding = 255
+ language = und
+ initializationData:
+ data = length 26, hash 4E58F6C7
+ sample 0:
+ time = 0
+ flags = 1
+ data = length 1706, hash F69B88EF
+ sample 1:
+ time = 21333
+ flags = 0
+ data = length 1705, hash 20F7927C
+ sample 2:
+ time = 42666
+ flags = 0
+ data = length 1900, hash D986BAA7
+ sample 3:
+ time = 64000
+ flags = 0
+ data = length 2224, hash DC056DA8
+ sample 4:
+ time = 85333
+ flags = 0
+ data = length 2157, hash D9587B29
+ sample 5:
+ time = 106666
+ flags = 0
+ data = length 2252, hash 9935FAC5
+ sample 6:
+ time = 128000
+ flags = 0
+ data = length 2025, hash 388EC449
+ sample 7:
+ time = 149333
+ flags = 0
+ data = length 1818, hash 148B1B72
+ sample 8:
+ time = 170666
+ flags = 0
+ data = length 1844, hash E997B535
+ sample 9:
+ time = 192000
+ flags = 0
+ data = length 1717, hash 2A9D93FB
+ sample 10:
+ time = 213333
+ flags = 0
+ data = length 1701, hash 40238DB5
+ sample 11:
+ time = 234666
+ flags = 0
+ data = length 1684, hash C46763BF
+ sample 12:
+ time = 256000
+ flags = 0
+ data = length 1747, hash 5FC3C86E
+ sample 13:
+ time = 277333
+ flags = 0
+ data = length 1750, hash 7F164780
+ sample 14:
+ time = 298666
+ flags = 0
+ data = length 1742, hash B853D24D
+ sample 15:
+ time = 320000
+ flags = 0
+ data = length 1839, hash FBFCDC4A
+ sample 16:
+ time = 341333
+ flags = 0
+ data = length 1809, hash C4722031
+ sample 17:
+ time = 362666
+ flags = 0
+ data = length 1746, hash 5990EC1F
+ sample 18:
+ time = 384000
+ flags = 0
+ data = length 1657, hash 1973E3C4
+ sample 19:
+ time = 405333
+ flags = 0
+ data = length 1862, hash 47500487
+ sample 20:
+ time = 426666
+ flags = 0
+ data = length 1687, hash 6789C2B4
+ sample 21:
+ time = 448000
+ flags = 0
+ data = length 1661, hash 26AB63E4
+ sample 22:
+ time = 469333
+ flags = 0
+ data = length 1671, hash 85AA94AD
+ sample 23:
+ time = 490666
+ flags = 0
+ data = length 1666, hash 60AF02C
+ sample 24:
+ time = 512000
+ flags = 0
+ data = length 1744, hash 93B89CC9
+ sample 25:
+ time = 533333
+ flags = 1
+ data = length 3498, hash 3013997F
+ sample 26:
+ time = 554666
+ flags = 0
+ data = length 1645, hash 912C7F11
+ sample 27:
+ time = 576000
+ flags = 0
+ data = length 1679, hash 5C1D5E4B
+ sample 28:
+ time = 597333
+ flags = 0
+ data = length 1653, hash AAFDF0C4
+ sample 29:
+ time = 618666
+ flags = 0
+ data = length 1793, hash 7DEDF0E7
+ sample 30:
+ time = 640000
+ flags = 0
+ data = length 1673, hash 5123A069
+ sample 31:
+ time = 661333
+ flags = 0
+ data = length 1631, hash 668D073B
+ sample 32:
+ time = 682666
+ flags = 0
+ data = length 1645, hash 4BA09D58
+ sample 33:
+ time = 704000
+ flags = 0
+ data = length 1715, hash 6696D08A
+ sample 34:
+ time = 725333
+ flags = 0
+ data = length 1913, hash F22841A5
+ sample 35:
+ time = 746666
+ flags = 0
+ data = length 1808, hash 3EB2EF9A
+ sample 36:
+ time = 768000
+ flags = 0
+ data = length 1703, hash A76E92E6
+ sample 37:
+ time = 789333
+ flags = 0
+ data = length 1663, hash D60564B6
+ sample 38:
+ time = 810666
+ flags = 0
+ data = length 1691, hash 3E9DB1D0
+ sample 39:
+ time = 832000
+ flags = 0
+ data = length 1682, hash 66EF509A
+ sample 40:
+ time = 853333
+ flags = 0
+ data = length 1661, hash 1F1114BE
+ sample 41:
+ time = 874666
+ flags = 0
+ data = length 1651, hash 37C9B7B6
+ sample 42:
+ time = 896000
+ flags = 0
+ data = length 1675, hash 94616355
+ sample 43:
+ time = 917333
+ flags = 0
+ data = length 1671, hash DA7F4549
+ sample 44:
+ time = 938666
+ flags = 0
+ data = length 1799, hash 49EF8B35
+ sample 45:
+ time = 960000
+ flags = 0
+ data = length 1807, hash C8BDF3C1
+ sample 46:
+ time = 981333
+ flags = 0
+ data = length 1674, hash 36EA2B2F
+ sample 47:
+ time = 1002666
+ flags = 0
+ data = length 1662, hash A865F92C
+ sample 48:
+ time = 1024000
+ flags = 0
+ data = length 1856, hash D20294BC
+ sample 49:
+ time = 1045333
+ flags = 0
+ data = length 1754, hash 54C7681A
+ sample 50:
+ time = 1066666
+ flags = 1
+ data = length 3408, hash 48774BB2
+ sample 51:
+ time = 1088000
+ flags = 0
+ data = length 1602, hash 8E895F43
+ sample 52:
+ time = 1109333
+ flags = 0
+ data = length 1624, hash 9E0AD8CF
+ sample 53:
+ time = 1130666
+ flags = 0
+ data = length 1616, hash 1F123433
+ sample 54:
+ time = 1152000
+ flags = 0
+ data = length 1671, hash 29644955
+ sample 55:
+ time = 1173333
+ flags = 0
+ data = length 1641, hash 55E8050C
+ sample 56:
+ time = 1194666
+ flags = 0
+ data = length 1670, hash CC133185
+ sample 57:
+ time = 1216000
+ flags = 536870912
+ data = length 1705, hash 35C3F104
+tracksEnded = true