Add JNI binding for speaker embedding extractor.

android/SherpaOnnxSpeakerIdentification/app/src/main/java/com/k2fsa/sherpa/onnx/speaker/identification/MainActivity.kt

@@ -58,6 +58,7 @@ class MainActivity : ComponentActivity() {
         }
 
         ActivityCompat.requestPermissions(this, permissions, REQUEST_RECORD_AUDIO_PERMISSION)
+
     }
 
     @Deprecated("Deprecated in Java")

android/SherpaOnnxSpeakerIdentification/app/src/main/java/com/k2fsa/sherpa/onnx/speaker/identification/Speaker.kt

@@ -0,0 +1,90 @@
+package com.k2fsa.sherpa.onnx
+
+import android.content.res.AssetManager
+
+
+data class SpeakerEmbeddingExtractorConfig(
+    val model: String,
+    var numThreads: Int = 1,
+    var debug: Boolean = false,
+    var provider: String = "cpu",
+)
+
+class SpeakerEmbeddingExtractorStream(var ptr: Long) {
+    fun acceptWaveform(samples: FloatArray, sampleRate: Int) = acceptWaveform(ptr, samples, sampleRate)
+
+    fun inputFinished() = inputFinished(ptr)
+
+    protected fun finalize() {
+        delete(ptr)
+        ptr = 0
+    }
+
+    private external fun myTest(ptr: Long, v: Array<FloatArray>)
+
+    fun release() = finalize()
+    private external fun acceptWaveform(ptr: Long, samples: FloatArray, sampleRate: Int)
+
+    private external fun inputFinished(ptr: Long)
+
+    private external fun delete(ptr: Long)
+    companion object {
+        init {
+            System.loadLibrary("sherpa-onnx-jni")
+        }
+    }
+}
+
+class SpeakerEmbeddingExtractor(
+    assetManager: AssetManager? = null,
+    config: SpeakerEmbeddingExtractorConfig,
+) {
+    private var ptr: Long
+
+    init {
+        ptr = if (assetManager != null) {
+            new(assetManager, config)
+        } else {
+            newFromFile(config)
+        }
+    }
+
+    protected fun finalize() {
+        delete(ptr)
+        ptr = 0
+    }
+
+    fun release() = finalize()
+
+    fun createStream(): SpeakerEmbeddingExtractorStream {
+        val p = createStream(ptr)
+        return SpeakerEmbeddingExtractorStream(p)
+    }
+
+    fun isReady(stream: SpeakerEmbeddingExtractorStream) = isReady(ptr, stream.ptr)
+    fun compute(stream: SpeakerEmbeddingExtractorStream) = compute(ptr, stream.ptr)
+
+    private external fun new(
+        assetManager: AssetManager,
+        config: SpeakerEmbeddingExtractorConfig,
+    ): Long
+
+    private external fun newFromFile(
+        config: SpeakerEmbeddingExtractorConfig,
+    ): Long
+
+    private external fun delete(ptr: Long)
+
+    private external fun createStream(ptr: Long): Long
+
+    private external fun isReady(ptr: Long, streamPtr: Long): Boolean
+
+
+    private external fun compute(ptr: Long, streamPtr: Long): FloatArray
+
+    companion object {
+        init {
+            System.loadLibrary("sherpa-onnx-jni")
+        }
+    }
+}

android/SherpaOnnxSpeakerIdentification/app/src/main/java/com/k2fsa/sherpa/onnx/speaker/identification/screens/Register.kt

@@ -1,6 +1,7 @@
 package com.k2fsa.sherpa.onnx.speaker.identification.screens
 
 import android.Manifest
+import android.annotation.SuppressLint
 import android.app.Activity
 import android.content.pm.PackageManager
 import android.media.AudioAttributes
@@ -20,6 +21,7 @@ import androidx.compose.foundation.layout.fillMaxWidth
 import androidx.compose.foundation.layout.padding
 import androidx.compose.foundation.layout.width
 import androidx.compose.material3.Button
+import androidx.compose.material3.MaterialTheme
 import androidx.compose.material3.OutlinedTextField
 import androidx.compose.material3.Text
 import androidx.compose.runtime.Composable
@@ -31,6 +33,7 @@ import androidx.compose.ui.Alignment
 import androidx.compose.ui.Modifier
 import androidx.compose.ui.platform.LocalContext
 import androidx.compose.ui.res.stringResource
+import androidx.compose.ui.text.font.FontWeight
 import androidx.compose.ui.tooling.preview.Preview
 import androidx.compose.ui.unit.dp
 import androidx.core.app.ActivityCompat
@@ -42,11 +45,26 @@ private var audioRecord: AudioRecord? = null
 
 private var sampleList: MutableList<FloatArray>? = null
 
+private var allSampleList: MutableList<MutableList<FloatArray>>? = null
+
+private var number = 0
+
+@SuppressLint("UnrememberedMutableState")
 @Preview
 @Composable
 fun RegisterScreen(modifier: Modifier = Modifier) {
     val activity = LocalContext.current as Activity
 
+    var firstTime by remember { mutableStateOf(true) }
+    if (firstTime) {
+        firstTime = false
+        // clear states
+
+        number = 0
+    }
+
+    var numberAudio by mutableStateOf(number)
+
     Box(
         modifier = Modifier.fillMaxSize(),
         contentAlignment = Alignment.TopCenter
@@ -108,6 +126,8 @@ fun RegisterScreen(modifier: Modifier = Modifier) {
                         }
 
                         Log.i(TAG, "Recording is stopped. ${sampleList?.count()}")
+
+                        ++number
                     }
                 }
             } else {
@@ -155,6 +175,12 @@ fun RegisterScreen(modifier: Modifier = Modifier) {
 
         Column(horizontalAlignment = Alignment.CenterHorizontally) {
             SpeakerNameRow(speakerName = speakerName, onValueChange = onSpeakerNameChange)
+            Text(
+                "Number of recordings: ${numberAudio}",
+                modifier = modifier.padding(24.dp),
+                style = MaterialTheme.typography.headlineMedium,
+                fontWeight = FontWeight.Bold,
+            )
             RegisterSpeakerButtonRow(
                 modifier,
                 isStarted = isStarted,
@@ -177,7 +203,9 @@ fun SpeakerNameRow(
             Text("Please input the speaker name")
         },
         singleLine = true,
-        modifier = modifier.fillMaxWidth().padding(8.dp)
+        modifier = modifier
+            .fillMaxWidth()
+            .padding(8.dp)
     )
 }
 

kotlin-api-examples/Main.kt

@@ -7,11 +7,43 @@ fun callback(samples: FloatArray): Unit {
 }
 
 fun main() {
+  testSpeakerRecognition()
   testTts()
   testAsr("transducer")
   testAsr("zipformer2-ctc")
 }
 
+fun computeEmbedding(extractor: SpeakerEmbeddingExtractor, filename: String): FloatArray {
+    var objArray = WaveReader.readWaveFromFile(
+        filename = filename,
+    )
+    var samples: FloatArray = objArray[0] as FloatArray
+    var sampleRate: Int = objArray[1] as Int
+
+    val stream = extractor.createStream()
+    stream.acceptWaveform(sampleRate = sampleRate, samples=samples)
+    stream.inputFinished()
+    check(extractor.isReady(stream))
+
+    val embedding = extractor.compute(stream)
+
+    stream.release()
+
+    return embedding
+}
+
+fun testSpeakerRecognition() {
+    val config = SpeakerEmbeddingExtractorConfig(
+        model="./3dspeaker_speech_eres2net_large_sv_zh-cn_3dspeaker_16k.onnx",
+        )
+    val extractor = SpeakerEmbeddingExtractor(config = config)
+
+    val embedding1a = computeEmbedding(extractor, "./speaker1_a_cn_16k.wav")
+    val embedding2a = computeEmbedding(extractor, "./speaker2_a_cn_16k.wav")
+    val embedding1b = computeEmbedding(extractor, "./speaker1_b_cn_16k.wav")
+    println(embedding1a.count())
+}
+
 fun testTts() {
   // see https://github.com/k2-fsa/sherpa-onnx/releases/tag/tts-models
   // https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-piper-en_US-amy-low.tar.bz2

kotlin-api-examples/Speaker.kt

@@ -0,0 +1 @@
+../android/SherpaOnnxSpeakerIdentification/app/src/main/java/com/k2fsa/sherpa/onnx/speaker/identification/Speaker.kt

kotlin-api-examples/run.sh

@@ -29,6 +29,22 @@ export LD_LIBRARY_PATH=$PWD/build/lib:$LD_LIBRARY_PATH
 
 cd ../kotlin-api-examples
 
+if [ ! -f ./3dspeaker_speech_eres2net_large_sv_zh-cn_3dspeaker_16k.onnx ]; then
+  wget -q https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-recongition-models/3dspeaker_speech_eres2net_large_sv_zh-cn_3dspeaker_16k.onnx
+fi
+
+if [ ! -f ./speaker1_a_cn_16k.wav ]; then
+  wget -q https://github.com/csukuangfj/sr-data/raw/main/test/3d-speaker/speaker1_a_cn_16k.wav
+fi
+
+if [ ! -f ./speaker1_b_cn_16k.wav ]; then
+  wget -q https://github.com/csukuangfj/sr-data/raw/main/test/3d-speaker/speaker1_b_cn_16k.wav
+fi
+
+if [ ! -f ./speaker2_a_cn_16k.wav ]; then
+  wget -q https://github.com/csukuangfj/sr-data/raw/main/test/3d-speaker/speaker2_a_cn_16k.wav
+fi
+
 if [ ! -f ./sherpa-onnx-streaming-zipformer-en-2023-02-21/tokens.txt ]; then
   git lfs install
   git clone https://huggingface.co/csukuangfj/sherpa-onnx-streaming-zipformer-en-2023-02-21
@@ -46,7 +62,7 @@ if [ ! -f ./vits-piper-en_US-amy-low/en_US-amy-low.onnx ]; then
   rm vits-piper-en_US-amy-low.tar.bz2
 fi
 
-kotlinc-jvm -include-runtime -d main.jar Main.kt WaveReader.kt SherpaOnnx.kt faked-asset-manager.kt Tts.kt
+kotlinc-jvm -include-runtime -d main.jar Main.kt WaveReader.kt SherpaOnnx.kt faked-asset-manager.kt Tts.kt Speaker.kt
 
 ls -lh main.jar