Add Audio tensor metadata

PiperOrigin-RevId: 364377123

tensorflow_lite_support/metadata/cc/metadata_version.cc

@@ -45,6 +45,7 @@ enum class SchemaMembers {
   kSubGraphMetadataInputTensorGroups = 5,
   kSubGraphMetadataOutputTensorGroups = 6,
   kProcessUnitOptionsRegexTokenizerOptions = 7,
+  kContentPropertiesAudioProperties = 8,
 };
 
 // Helper class to compare semantic versions in terms of three integers, major,
@@ -107,6 +108,8 @@ Version GetMemberVersion(SchemaMembers member) {
       return Version(1, 2, 0);
     case SchemaMembers::kProcessUnitOptionsRegexTokenizerOptions:
       return Version(1, 2, 1);
+    case SchemaMembers::kContentPropertiesAudioProperties:
+      return Version(1, 3, 0);
     default:
       // Should never happen.
       TFLITE_LOG(FATAL) << "Unsupported schema member: "
@@ -176,6 +179,19 @@ void UpdateMinimumVersionForTable<tflite::ProcessUnit>(
   }
 }
 
+template <>
+void UpdateMinimumVersionForTable<tflite::Content>(const tflite::Content* table,
+                                                   Version* min_version) {
+  if (table == nullptr) return;
+
+  // Checks the ContenProperties field.
+  if (table->content_properties_type() == ContentProperties_AudioProperties) {
+    UpdateMinimumVersion(
+        GetMemberVersion(SchemaMembers::kContentPropertiesAudioProperties),
+        min_version);
+  }
+}
+
 template <>
 void UpdateMinimumVersionForTable<tflite::TensorMetadata>(
     const tflite::TensorMetadata* table, Version* min_version) {
@@ -188,6 +204,9 @@ void UpdateMinimumVersionForTable<tflite::TensorMetadata>(
   // Checks the process_units field.
   UpdateMinimumVersionForArray<tflite::ProcessUnit>(table->process_units(),
                                                     min_version);
+
+  // Check the content field.
+  UpdateMinimumVersionForTable<tflite::Content>(table->content(), min_version);
 }
 
 template <>

tensorflow_lite_support/metadata/java/src/java/org/tensorflow/lite/support/metadata/MetadataParser.java

@@ -21,7 +21,7 @@ public final class MetadataParser {
    * The version of the metadata parser that this metadata extractor library is depending on. The
    * value should match the value of "Schema Semantic version" in metadata_schema.fbs.
    */
-  public static final String VERSION = "1.2.1";
+  public static final String VERSION = "1.3.0";
 
   private MetadataParser() {}
 }

tensorflow_lite_support/metadata/metadata_schema.fbs

@@ -50,7 +50,7 @@ namespace tflite;
 // for which they were added.
 //
 // LINT.IfChange
-// Schema Semantic version: 1.2.1
+// Schema Semantic version: 1.3.0
 // LINT.ThenChange(//tensorflow_lite_support/\
 //     metadata/java/src/java/org/tensorflow/lite/support/metadata/\
 //     MetadataParser.java)
@@ -69,6 +69,7 @@ file_identifier "M001";
 // 1.2.0 - Added input_tensor_group to SubGraphMetadata.
 //         Added output_tensor_group to SubGraphMetadata.
 // 1.2.1 - Added RegexTokenizerOptions to ProcessUnitOptions.
+// 1.3.0 - Added AudioProperties to ContentProperties.
 
 // File extension of any written files.
 file_extension "tflitemeta";
@@ -232,6 +233,20 @@ enum BoundingBoxType : byte {
 
 }
 
+// The properties for audio tensors.
+// Added in: 1.3.0
+table AudioProperties {
+  // The sample rate in Hz when the audio was captured.
+  sample_rate:uint;
+
+  // The channel count of the audio.
+  channels:uint;
+
+  // The minimum required number of samples in order to run inference properly.
+  // Optional for fixed-size audio tensors.
+  min_required_samples:uint;
+}
+
 enum CoordinateType : byte {
   // The coordinates are float values from 0 to 1.
   RATIO = 0,
@@ -267,6 +282,8 @@ union ContentProperties {
   FeatureProperties,
   ImageProperties,
   BoundingBoxProperties,
+  // Added in: 1.3.0
+  AudioProperties,
 }
 
 table ValueRange {