From 757a44b11632c83db1470cd6620ea7d180e518ba Mon Sep 17 00:00:00 2001
From: Fangjun Kuang <csukuangfj@gmail.com>
Date: Sun, 23 Jun 2024 18:34:18 +0800
Subject: [PATCH] Add VAD + microphone example for Java API. (#1045)

---
 java-api-examples/README.md           |  6 ++
 java-api-examples/VadFromMic.java     | 98 +++++++++++++++++++++++++++
 java-api-examples/run-vad-from-mic.sh | 35 ++++++++++
 3 files changed, 139 insertions(+)
 create mode 100644 java-api-examples/VadFromMic.java
 create mode 100755 java-api-examples/run-vad-from-mic.sh

diff --git a/java-api-examples/README.md b/java-api-examples/README.md
index 18f53fae..420fdb71 100755
--- a/java-api-examples/README.md
+++ b/java-api-examples/README.md
@@ -57,6 +57,12 @@ The punctuation model supports both English and Chinese.
 ./run-speaker-identification.sh
 ```
 
+## VAD with a microphone
+
+```bash
+./run-vad-from-mic.sh
+```
+
 ## VAD (Remove silence)
 
 ```bash
diff --git a/java-api-examples/VadFromMic.java b/java-api-examples/VadFromMic.java
new file mode 100644
index 00000000..95fd3b19
--- /dev/null
+++ b/java-api-examples/VadFromMic.java
@@ -0,0 +1,98 @@
+// Copyright 2024 Xiaomi Corporation
+
+// This file shows how to use a silero_vad model to detect speech
+// and save detected speech into a wave file.
+
+import com.k2fsa.sherpa.onnx.*;
+import javax.sound.sampled.*;
+
+public class VadFromMic {
+  public static void main(String[] args) {
+    int sampleRate = 16000;
+    int windowSize = 512;
+    // please download ./silero_vad.onnx from
+    // https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
+    String model = "./silero_vad.onnx";
+    SileroVadModelConfig sileroVad =
+        SileroVadModelConfig.builder()
+            .setModel(model)
+            .setThreshold(0.5f)
+            .setMinSilenceDuration(0.25f)
+            .setMinSpeechDuration(0.5f)
+            .setWindowSize(windowSize)
+            .build();
+
+    VadModelConfig config =
+        VadModelConfig.builder()
+            .setSileroVadModelConfig(sileroVad)
+            .setSampleRate(sampleRate)
+            .setNumThreads(1)
+            .setDebug(true)
+            .setProvider("cpu")
+            .build();
+
+    Vad vad = new Vad(config);
+
+    // https://docs.oracle.com/javase/8/docs/api/javax/sound/sampled/AudioFormat.html
+    // Linear PCM, 16000Hz, 16-bit, 1 channel, signed, little endian
+    AudioFormat format = new AudioFormat(sampleRate, 16, 1, true, false);
+
+    // https://docs.oracle.com/javase/8/docs/api/javax/sound/sampled/DataLine.Info.html#Info-java.lang.Class-javax.sound.sampled.AudioFormat-int-
+    DataLine.Info info = new DataLine.Info(TargetDataLine.class, format);
+    TargetDataLine targetDataLine;
+    try {
+      targetDataLine = (TargetDataLine) AudioSystem.getLine(info);
+      targetDataLine.open(format);
+      targetDataLine.start();
+    } catch (LineUnavailableException e) {
+      System.out.println("Failed to open target data line: " + e.getMessage());
+      vad.release();
+      return;
+    }
+
+    boolean printed = false;
+    int index = 0;
+
+    byte[] buffer = new byte[windowSize * 2];
+    float[] samples = new float[windowSize];
+
+    while (targetDataLine.isOpen()) {
+      int n = targetDataLine.read(buffer, 0, buffer.length);
+      if (n <= 0) {
+        System.out.printf("Got %d bytes. Expected %d bytes.\n", n, buffer.length);
+        continue;
+      }
+      for (int i = 0; i != windowSize; ++i) {
+        short low = buffer[2 * i];
+        short high = buffer[2 * i + 1];
+        int s = (high << 8) + low;
+        samples[i] = (float) s / 32768;
+      }
+
+      vad.acceptWaveform(samples);
+      if (vad.isSpeechDetected() && !printed) {
+        System.out.println("Detected speech");
+        printed = true;
+      }
+
+      if (!vad.isSpeechDetected()) {
+        printed = false;
+      }
+
+      while (!vad.empty()) {
+        float[] segment = vad.front().getSamples();
+        float duration = segment.length / (float) sampleRate;
+        System.out.printf("Duration: %.3f seconds\n", duration);
+
+        String filename = String.format("seg-%d-%.3fs.wav", index, duration);
+        index += 1;
+        WaveWriter.write(filename, segment, sampleRate);
+        System.out.printf("Saved to %s\n", filename);
+        System.out.println("----------");
+        vad.pop();
+      }
+    }
+
+    vad.release();
+  }
+}
diff --git a/java-api-examples/run-vad-from-mic.sh b/java-api-examples/run-vad-from-mic.sh
new file mode 100755
index 00000000..320e97cb
--- /dev/null
+++ b/java-api-examples/run-vad-from-mic.sh
@@ -0,0 +1,35 @@
+#!/usr/bin/env bash
+
+set -ex
+
+if [[ ! -f ../build/lib/libsherpa-onnx-jni.dylib  && ! -f ../build/lib/libsherpa-onnx-jni.so ]]; then
+  mkdir -p ../build
+  pushd ../build
+  cmake \
+    -DSHERPA_ONNX_ENABLE_PYTHON=OFF \
+    -DSHERPA_ONNX_ENABLE_TESTS=OFF \
+    -DSHERPA_ONNX_ENABLE_CHECK=OFF \
+    -DBUILD_SHARED_LIBS=ON \
+    -DSHERPA_ONNX_ENABLE_PORTAUDIO=OFF \
+    -DSHERPA_ONNX_ENABLE_JNI=ON \
+    ..
+
+  make -j4
+  ls -lh lib
+  popd
+fi
+
+if [ ! -f ../sherpa-onnx/java-api/build/sherpa-onnx.jar ]; then
+  pushd ../sherpa-onnx/java-api
+  make
+  popd
+fi
+
+if [ ! -f ./silero_vad.onnx ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
+fi
+
+java \
+  -Djava.library.path=$PWD/../build/lib \
+  -cp ../sherpa-onnx/java-api/build/sherpa-onnx.jar \
+  ./VadFromMic.java