Fix ISO 14496-3 table to use 24kHz for AAC instead of 23kHz.

- 23kHz must have been incorrectly copied and propagated. The source ISO 14496-3 table uses 24kHz in this slot. - The old value is simply wrong, and the enums map directly to the table (offset by 1). Force rename the typos to represent the correct value. PiperOrigin-RevId: 713774201
AOMediaCodec · Jan 9, 2025 · d4fcda0 · d4fcda0
1 parent 810dc05
commit d4fcda0
Show file tree

Hide file tree

Showing 7 changed files with 64 additions and 6 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -57,6 +57,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
     annotations.
 -   Fix compliance with ISO 14496-1:2010 when writing AAC Codec Config OBUs
     (AOMediaCodec/libiamf#119).
+-   Fix issues when using AAC with a 24 kHz sample rate.
 -   Permit one fully trimmed audio frame at the end of a substream.
 
 ## [1.0.0] - 2024-01-26

diff --git a/iamf/cli/lookup_tables.h b/iamf/cli/lookup_tables.h
@@ -109,7 +109,7 @@ class LookupTables {
          {AAC_SAMPLE_FREQUENCY_INDEX_48000, k48000},
          {AAC_SAMPLE_FREQUENCY_INDEX_44100, k44100},
          {AAC_SAMPLE_FREQUENCY_INDEX_32000, k32000},
-         {AAC_SAMPLE_FREQUENCY_INDEX_23000, k23000},
+         {AAC_SAMPLE_FREQUENCY_INDEX_24000, k24000},
          {AAC_SAMPLE_FREQUENCY_INDEX_22050, k22050},
          {AAC_SAMPLE_FREQUENCY_INDEX_16000, k16000},
          {AAC_SAMPLE_FREQUENCY_INDEX_12000, k12000},

diff --git a/iamf/cli/proto/codec_config.proto b/iamf/cli/proto/codec_config.proto
@@ -71,7 +71,7 @@ enum SampleFrequencyIndex {
   AAC_SAMPLE_FREQUENCY_INDEX_48000 = 4;
   AAC_SAMPLE_FREQUENCY_INDEX_44100 = 5;
   AAC_SAMPLE_FREQUENCY_INDEX_32000 = 6;
-  AAC_SAMPLE_FREQUENCY_INDEX_23000 = 7;
+  AAC_SAMPLE_FREQUENCY_INDEX_24000 = 7;
   AAC_SAMPLE_FREQUENCY_INDEX_22050 = 8;
   AAC_SAMPLE_FREQUENCY_INDEX_16000 = 9;
   AAC_SAMPLE_FREQUENCY_INDEX_12000 = 10;

diff --git a/iamf/cli/proto_to_obu/tests/codec_config_generator_test.cc b/iamf/cli/proto_to_obu/tests/codec_config_generator_test.cc
@@ -796,6 +796,26 @@ TEST_F(CodecConfigGeneratorTest, InvalidUnknownSamplingFrequencyIndex) {
   EXPECT_FALSE(InitAndGenerate().ok());
 }
 
+TEST_F(CodecConfigGeneratorTest, ConfiguresAacWithImplicitSamplingFrequency) {
+  InitMetadataForAac(codec_config_metadata_);
+  codec_config_metadata_.at(0)
+      .mutable_codec_config()
+      ->mutable_decoder_config_aac()
+      ->mutable_decoder_specific_info()
+      ->set_sample_frequency_index(
+          iamf_tools_cli_proto::AAC_SAMPLE_FREQUENCY_INDEX_24000);
+
+  const auto output_obus = InitAndGenerate();
+  ASSERT_THAT(output_obus, IsOk());
+
+  const auto& audio_specific_config =
+      std::get<AacDecoderConfig>(
+          output_obus->at(kCodecConfigId).GetCodecConfig().decoder_config)
+          .decoder_specific_info_.audio_specific_config;
+  EXPECT_EQ(audio_specific_config.sample_frequency_index_,
+            AudioSpecificConfig::SampleFrequencyIndex::k24000);
+}
+
 TEST_F(CodecConfigGeneratorTest, ConfiguresAacWithExplicitSamplingFrequency) {
   InitMetadataForAac(codec_config_metadata_);
   codec_config_metadata_.at(0)

diff --git a/iamf/obu/decoder_config/aac_decoder_config.cc b/iamf/obu/decoder_config/aac_decoder_config.cc
@@ -279,7 +279,7 @@ absl::Status AacDecoderConfig::GetOutputSampleRate(
                                               {k48000, 48000},
                                               {k44100, 44100},
                                               {k32000, 32000},
-                                              {k23000, 23000},
+                                              {k24000, 24000},
                                               {k22050, 22050},
                                               {k16000, 16000},
                                               {k12000, 12000},

diff --git a/iamf/obu/decoder_config/aac_decoder_config.h b/iamf/obu/decoder_config/aac_decoder_config.h
@@ -35,7 +35,7 @@ class AudioSpecificConfig {
     k48000 = 3,
     k44100 = 4,
     k32000 = 5,
-    k23000 = 6,
+    k24000 = 6,
     k22050 = 7,
     k16000 = 8,
     k12000 = 9,

diff --git a/iamf/obu/decoder_config/tests/aac_decoder_config_test.cc b/iamf/obu/decoder_config/tests/aac_decoder_config_test.cc
@@ -40,6 +40,13 @@ constexpr uint8_t kUpperByteSerializedSamplingFrequencyIndex64000 =
 constexpr uint8_t kLowerByteSerializedSamplingFrequencyIndex64000 =
     (static_cast<uint8_t>(SampleFrequencyIndex::k64000) & 0x01) << 7;
 
+// Despite being represented in 4-bits the AAC Sampling Frequency Index 24000 is
+// serialized across a byte boundary.
+constexpr uint8_t kUpperByteSerializedSamplingFrequencyIndex24000 =
+    (static_cast<uint8_t>(SampleFrequencyIndex::k24000) & 0x0e) >> 1;
+constexpr uint8_t kLowerByteSerializedSamplingFrequencyIndex24000 =
+    (static_cast<uint8_t>(SampleFrequencyIndex::k24000) & 0x01) << 7;
+
 constexpr uint8_t kUpperByteSerializedSamplingFrequencyIndexEscape =
     (static_cast<uint8_t>(SampleFrequencyIndex::kEscapeValue) & 0x0e) >> 1;
 constexpr uint8_t kLowerByteSerializedSamplingFrequencyIndexEscape =
@@ -185,7 +192,7 @@ TEST(AacDecoderConfig, ValidatesExtensionFlag) {
   EXPECT_FALSE(aac_decoder_config.Validate().ok());
 }
 
-TEST(AudioSpecificConfig, ReadsWithImplicitSampleFrequency) {
+TEST(AudioSpecificConfig, ReadsWithImplicitSampleFrequency64000) {
   std::vector<uint8_t> data = {
       // `audio_object_type`, upper 3 bits of `sample_frequency_index`.
       AudioSpecificConfig::kAudioObjectType << 3 |
@@ -215,6 +222,25 @@ TEST(AudioSpecificConfig, ReadsWithImplicitSampleFrequency) {
             AudioSpecificConfig::GaSpecificConfig::kExtensionFlag);
 }
 
+TEST(AudioSpecificConfig, ReadsWithImplicitSampleFrequency24000) {
+  const std::vector<uint8_t> data = {
+      // `audio_object_type`, upper 3 bits of `sample_frequency_index`.
+      AudioSpecificConfig::kAudioObjectType << 3 |
+          kUpperByteSerializedSamplingFrequencyIndex24000,
+      // lower bit of `sample_frequency_index`,
+      // `channel_configuration`, `frame_length_flag`,
+      // `depends_on_core_coder`, `extension_flag`.
+      kLowerByteSerializedSamplingFrequencyIndex24000 |
+          kChannelConfigurationAndGaSpecificConfigMask};
+  AudioSpecificConfig audio_specific_config;
+  auto rb = MemoryBasedReadBitBuffer::CreateFromSpan(1024, data);
+
+  EXPECT_THAT(audio_specific_config.Read(*rb), IsOk());
+
+  EXPECT_EQ(audio_specific_config.sample_frequency_index_,
+            SampleFrequencyIndex::k24000);
+}
+
 TEST(AudioSpecificConfig, ReadsWithExplicitSampleFrequency) {
   constexpr uint32_t kSampleFrequency = 48000;
   std::vector<uint8_t> data = {
@@ -688,7 +714,7 @@ TEST_F(AacTest, OverflowBufferSizeDbOver24Bits) {
   TestWriteDecoderConfig();
 }
 
-TEST_F(AacTest, GetImplicitSampleRate) {
+TEST_F(AacTest, GetImplicitSampleRate64000) {
   aac_decoder_config_.decoder_specific_info_.audio_specific_config
       .sample_frequency_index_ = SampleFrequencyIndex::k64000;
 
@@ -699,6 +725,17 @@ TEST_F(AacTest, GetImplicitSampleRate) {
   EXPECT_EQ(output_sample_rate, 64000);
 }
 
+TEST_F(AacTest, GetImplicitSampleRate24000) {
+  aac_decoder_config_.decoder_specific_info_.audio_specific_config
+      .sample_frequency_index_ = SampleFrequencyIndex::k24000;
+
+  uint32_t output_sample_rate;
+  EXPECT_THAT(aac_decoder_config_.GetOutputSampleRate(output_sample_rate),
+              IsOk());
+
+  EXPECT_EQ(output_sample_rate, 24000);
+}
+
 TEST_F(AacTest, GetExplicitSampleRate) {
   aac_decoder_config_.decoder_specific_info_.audio_specific_config
       .sample_frequency_index_ = SampleFrequencyIndex::kEscapeValue;