Add JavaScript (node-addon) API for ten-vad (#2383)

This PR adds support for the new ten-vad model in both the Node.js addon examples and the HarmonyOS wrapper.

- Introduce TenVadConfig alongside existing SileroVadConfig and extend the VadConfig API.
- Update C++ addon to parse ten-vad parameters and pass them through to the detector.
- Modify Node.js example scripts to let users switch between silero and ten-vad and to normalize generated filenames.

 export { listRawfileDir, readWave, readWaveFromBinary, } from "libsherpa_onnx.so";
 
-export { CircularBuffer, SileroVadConfig, SpeechSegment, Vad, VadConfig, } from './src/main/ets/components/Vad';
+export { CircularBuffer, SileroVadConfig, TenVadConfig, SpeechSegment, Vad, VadConfig, } from './src/main/ets/components/Vad';
 
 
 export { Samples,

@@ -294,6 +294,25 @@ static SherpaOnnxSileroVadModelConfig GetSileroVadConfig(
   return c;
 }
 
+static SherpaOnnxTenVadModelConfig GetTenVadConfig(const Napi::Object &obj) {
+  SherpaOnnxTenVadModelConfig c;
+  memset(&c, 0, sizeof(c));
+
+  if (!obj.Has("tenVad") || !obj.Get("tenVad").IsObject()) {
+    return c;
+  }
+
+  Napi::Object o = obj.Get("tenVad").As<Napi::Object>();
+  SHERPA_ONNX_ASSIGN_ATTR_STR(model, model);
+  SHERPA_ONNX_ASSIGN_ATTR_FLOAT(threshold, threshold);
+  SHERPA_ONNX_ASSIGN_ATTR_FLOAT(min_silence_duration, minSilenceDuration);
+  SHERPA_ONNX_ASSIGN_ATTR_FLOAT(min_speech_duration, minSpeechDuration);
+  SHERPA_ONNX_ASSIGN_ATTR_INT32(window_size, windowSize);
+  SHERPA_ONNX_ASSIGN_ATTR_FLOAT(max_speech_duration, maxSpeechDuration);
+
+  return c;
+}
+
 static Napi::External<SherpaOnnxVoiceActivityDetector>
 CreateVoiceActivityDetectorWrapper(const Napi::CallbackInfo &info) {
   Napi::Env env = info.Env();
@@ -339,6 +358,7 @@ CreateVoiceActivityDetectorWrapper(const Napi::CallbackInfo &info) {
   SherpaOnnxVadModelConfig c;
   memset(&c, 0, sizeof(c));
   c.silero_vad = GetSileroVadConfig(o);
+  c.ten_vad = GetTenVadConfig(o);
 
   SHERPA_ONNX_ASSIGN_ATTR_INT32(sample_rate, sampleRate);
   SHERPA_ONNX_ASSIGN_ATTR_INT32(num_threads, numThreads);
@@ -369,6 +389,7 @@ CreateVoiceActivityDetectorWrapper(const Napi::CallbackInfo &info) {
       SherpaOnnxCreateVoiceActivityDetector(&c, buffer_size_in_seconds);
 #endif
   SHERPA_ONNX_DELETE_C_STR(c.silero_vad.model);
+  SHERPA_ONNX_DELETE_C_STR(c.ten_vad.model);
   SHERPA_ONNX_DELETE_C_STR(c.provider);
 
   return Napi::External<SherpaOnnxVoiceActivityDetector>::New(

@@ -23,25 +23,48 @@ export class SileroVadConfig {
   public minSpeechDuration: number;
   public minSilenceDuration: number;
   public windowSize: number;
+  public maxSpeechDuration: number;
 
   public constructor(model: string, threshold: number, minSpeechDuration: number, minSilenceDuration: number,
-    windowSize: number) {
+    windowSize: number, maxSpeechDuration: number = 20) {
     this.model = model;
     this.threshold = threshold;
     this.minSpeechDuration = minSpeechDuration;
     this.minSilenceDuration = minSilenceDuration;
     this.windowSize = windowSize;
+    this.maxSpeechDuration = maxSpeechDuration
+  }
+}
+
+export class TenVadConfig {
+  public model: string;
+  public threshold: number;
+  public minSpeechDuration: number;
+  public minSilenceDuration: number;
+  public windowSize: number;
+  public maxSpeechDuration: number;
+
+  public constructor(model: string, threshold: number, minSpeechDuration: number, minSilenceDuration: number,
+    windowSize: number, maxSpeechDuration: number = 20) {
+    this.model = model;
+    this.threshold = threshold;
+    this.minSpeechDuration = minSpeechDuration;
+    this.minSilenceDuration = minSilenceDuration;
+    this.windowSize = windowSize;
+    this.maxSpeechDuration = maxSpeechDuration
   }
 }
 
 export class VadConfig {
   public sileroVad: SileroVadConfig;
+  public tenVad: TenVadConfig;
   public sampleRate: number;
   public debug: boolean;
   public numThreads: number;
 
-  public constructor(sileroVad: SileroVadConfig, sampleRate: number, debug: boolean, numThreads: number) {
+  public constructor(sileroVad: SileroVadConfig, tenVad: TenVadConfig, sampleRate: number, debug: boolean, numThreads: number) {
     this.sileroVad = sileroVad;
+    this.tenVad = tenVad;
     this.sampleRate = sampleRate;
     this.debug = debug;
     this.numThreads = numThreads;
@@ -127,4 +150,4 @@ export class Vad {
   flush(): void {
     voiceActivityDetectorFlush(this.handle);
   }
-}
\ No newline at end of file
+}

@@ -6,6 +6,7 @@ import {
   OnlineRecognizerResult,
   readWaveFromBinary,
   SileroVadConfig,
+  TenVadConfig,
   SpeechSegment,
   Vad,
   VadConfig,
@@ -31,6 +32,13 @@ function initVad(context: Context): Vad {
       0.5,
       512,
     ),
+    new TenVadConfig(
+      '', // set it to ten-vad.onnx to use ten-vad
+      0.5,
+      0.25,
+      0.5,
+      256,
+    ),
     16000,
     true,
     1,
@@ -93,7 +101,12 @@ function decodeFile(filename: string): string {
   console.log(`samples length ${wave.samples.length}`);
   const resultList: string[] = [];
 
-  const windowSize: number = vad.config.sileroVad.windowSize;
+  let windowSize: number = vad.config.sileroVad.windowSize;
+
+  if (vad.config.tenVad.model != '') {
+    windowSize = vad.config.tenVad.windowSize;
+  }
+
   for (let i = 0; i < wave.samples.length; i += windowSize) {
     const thisWindow: Float32Array = wave.samples.subarray(i, i + windowSize)
     vad.acceptWaveform(thisWindow);
@@ -138,7 +151,12 @@ function decodeFile(filename: string): string {
 function decodeMic(samples: Float32Array) {
   const resultList: string[] = [];
 
-  const windowSize: number = vad.config.sileroVad.windowSize;
+  let windowSize: number = vad.config.sileroVad.windowSize;
+
+  if (vad.config.tenVad.model != '') {
+    windowSize = vad.config.tenVad.windowSize;
+  }
+
   for (let i = 0; i < samples.length; i += windowSize) {
     const thisWindow: Float32Array = samples.subarray(i, i + windowSize)
     vad.acceptWaveform(thisWindow);

@@ -97,9 +97,11 @@ ai.on('data', data => {
       console.log(`${index}: ${text}`);
 
       const filename = `${index}-${text}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
+
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});

@@ -94,9 +94,11 @@ ai.on('data', data => {
       console.log(`${index}: ${text}`);
 
       const filename = `${index}-${text}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
+
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});

@@ -92,9 +92,11 @@ ai.on('data', data => {
       console.log(`${index}: ${text}`);
 
       const filename = `${index}-${text}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
+
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});

@@ -95,9 +95,11 @@ ai.on('data', data => {
       console.log(`${index}: ${text}`);
 
       const filename = `${index}-${text}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
+
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});

@@ -97,9 +97,11 @@ ai.on('data', data => {
       console.log(`${index}: ${text}`);
 
       const filename = `${index}-${text}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
+
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});

@@ -93,9 +93,11 @@ ai.on('data', data => {
       console.log(`${index}: ${text}`);
 
       const filename = `${index}-${text}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
+
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});

@@ -93,9 +93,11 @@ ai.on('data', data => {
       console.log(`${index}: ${text}`);
 
       const filename = `${index}-${text}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
+
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});

@@ -8,14 +8,28 @@ const sherpa_onnx = require('sherpa-onnx-node');
 function createVad() {
   // please download silero_vad.onnx from
   // https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
+  //
+  // OR
+  //
+  // please download ten-vad.onnx from
+  // https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/ten-vad.onnx
   const config = {
     sileroVad: {
+      // model: '',
       model: './silero_vad.onnx',
       threshold: 0.5,
       minSpeechDuration: 0.25,
       minSilenceDuration: 0.5,
       windowSize: 512,
     },
+    tenVad: {
+      model: '',
+      // model: './ten-vad.onnx',
+      threshold: 0.5,
+      minSpeechDuration: 0.25,
+      minSilenceDuration: 0.5,
+      windowSize: 256,
+    },
     sampleRate: 16000,
     debug: true,
     numThreads: 1,
@@ -47,7 +61,10 @@ const ai = new portAudio.AudioIO({
 let printed = false;
 let index = 0;
 ai.on('data', data => {
-  const windowSize = vad.config.sileroVad.windowSize;
+  const windowSize = vad.config.sileroVad.model != '' ?
+      vad.config.sileroVad.windowSize :
+      vad.config.tenVad.windowSize;
+
   buffer.push(new Float32Array(data.buffer));
   while (buffer.size() > windowSize) {
     const samples = buffer.get(buffer.head(), windowSize);
@@ -66,9 +83,10 @@ ai.on('data', data => {
       const segment = vad.front();
       vad.pop();
       const filename = `${index}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});

@@ -91,9 +91,11 @@ ai.on('data', data => {
       const fullLang = display.of(lang);
 
       const filename = `${index}-${fullLang}-${
-          new Date()
-              .toLocaleTimeString('en-US', {hour12: false})
-              .split(' ')[0]}.wav`;
+                           new Date()
+                               .toLocaleTimeString('en-US', {hour12: false})
+                               .split(' ')[0]}.wav`
+                           .replace(/:/g, '-');
+
       sherpa_onnx.writeWave(
           filename,
           {samples: segment.samples, sampleRate: vad.config.sampleRate});