sherpa-onnx-node 1.0.20 → 1.0.22

package/audio-tagg.js

@@ -0,0 +1,26 @@
+const addon = require('./addon.js');
+const non_streaming_asr = require('./non-streaming-asr.js');
+
+class AudioTagging {
+  constructor(config) {
+    this.handle = addon.createAudioTagging(config);
+    this.config = config;
+  }
+
+  createStream() {
+    return new non_streaming_asr.OfflineStream(
+        addon.audioTaggingCreateOfflineStream(this.handle));
+  }
+
+  /* Return an array. Each element is
+   * an object {name: "xxx", prob: xxx, index: xxx};
+   *
+   */
+  compute(stream, topK = -1) {
+    return addon.audioTaggingCompute(this.handle, stream.handle, topK);
+  }
+}
+
+module.exports = {
+  AudioTagging,
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "sherpa-onnx-node",
-  "version": "1.0.20",
+  "version": "1.0.22",
   "description": "Speech-to-text and text-to-speech using Next-gen Kaldi without internet connection",
   "main": "sherpa-onnx.js",
   "scripts": {
@@ -45,10 +45,10 @@
   },
   "homepage": "https://github.com/csukuangfj/sherpa-onnx#readme",
   "optionalDependencies": {
-    "sherpa-onnx-darwin-arm64": "^1.0.20",
-    "sherpa-onnx-darwin-x64": "^1.0.20",
-    "sherpa-onnx-linux-x64": "^1.0.20",
-    "sherpa-onnx-linux-arm64": "^1.0.20",
-    "sherpa-onnx-win-x64": "^1.0.20"
+    "sherpa-onnx-darwin-arm64": "^1.0.22",
+    "sherpa-onnx-darwin-x64": "^1.0.22",
+    "sherpa-onnx-linux-x64": "^1.0.22",
+    "sherpa-onnx-linux-arm64": "^1.0.22",
+    "sherpa-onnx-win-x64": "^1.0.22"
   }
 }

package/sherpa-onnx.js

@@ -4,6 +4,8 @@ const non_streaming_asr = require('./non-streaming-asr.js');
 const non_streaming_tts = require('./non-streaming-tts.js');
 const vad = require('./vad.js');
 const slid = require('./spoken-language-identification.js');
+const sid = require('./speaker-identification.js');
+const at = require('./audio-tagg.js');
 
 module.exports = {
   OnlineRecognizer: streaming_asr.OnlineRecognizer,
@@ -15,4 +17,7 @@ module.exports = {
   Vad: vad.Vad,
   CircularBuffer: vad.CircularBuffer,
   SpokenLanguageIdentification: slid.SpokenLanguageIdentification,
+  SpeakerEmbeddingExtractor: sid.SpeakerEmbeddingExtractor,
+  SpeakerEmbeddingManager: sid.SpeakerEmbeddingManager,
+  AudioTagging: at.AudioTagging,
 }

package/speaker-identification.js

@@ -0,0 +1,102 @@
+const addon = require('./addon.js');
+const streaming_asr = require('./streaming-asr.js');
+
+class SpeakerEmbeddingExtractor {
+  constructor(config) {
+    this.handle = addon.createSpeakerEmbeddingExtractor(config);
+    this.config = config;
+    this.dim = addon.speakerEmbeddingExtractorDim(this.handle);
+  }
+
+  createStream() {
+    return new streaming_asr.OnlineStream(
+        addon.speakerEmbeddingExtractorCreateStream(this.handle));
+  }
+
+  isReady(stream) {
+    return addon.speakerEmbeddingExtractorIsReady(this.handle, stream.handle);
+  }
+
+  // return a float32 array
+  compute(stream) {
+    return addon.speakerEmbeddingExtractorComputeEmbedding(
+        this.handle, stream.handle);
+  }
+}
+
+function flatten(arrayList) {
+  let n = 0;
+  for (let i = 0; i < arrayList.length; ++i) {
+    n += arrayList[i].length;
+  }
+  let ans = new Float32Array(n);
+
+  let offset = 0;
+  for (let i = 0; i < arrayList.length; ++i) {
+    ans.set(arrayList[i], offset);
+    offset += arrayList[i].length;
+  }
+  return ans;
+}
+
+class SpeakerEmbeddingManager {
+  constructor(dim) {
+    this.handle = addon.createSpeakerEmbeddingManager(dim);
+    this.dim = dim;
+  }
+
+  /*
+   obj = {name: "xxx", v: a-float32-array}
+   */
+  add(obj) {
+    return addon.speakerEmbeddingManagerAdd(this.handle, obj);
+  }
+
+  /*
+   * obj =
+   * {name: "xxx", v: [float32_array1, float32_array2, ..., float32_arrayn]
+   */
+  addMulti(obj) {
+    const c = {
+      name: obj.name,
+      vv: flatten(obj.v),
+      n: obj.v.length,
+    };
+    return addon.speakerEmbeddingManagerAddListFlattened(this.handle, c);
+  }
+
+  remove(name) {
+    return addon.speakerEmbeddingManagerRemove(this.handle, name);
+  }
+
+  /*
+   * obj = {v: a-float32-array, threshold: a-float }
+   */
+  search(obj) {
+    return addon.speakerEmbeddingManagerSearch(this.handle, obj);
+  }
+
+  /*
+   * obj = {name: 'xxx', v: a-float32-array, threshold: a-float }
+   */
+  verify(obj) {
+    return addon.speakerEmbeddingManagerVerify(this.handle, obj);
+  }
+
+  contains(name) {
+    return addon.speakerEmbeddingManagerContains(this.handle, name);
+  }
+
+  getNumSpeakers() {
+    return addon.speakerEmbeddingManagerNumSpeakers(this.handle);
+  }
+
+  getAllSpeakerNames() {
+    return addon.speakerEmbeddingManagerGetAllSpeakers(this.handle);
+  }
+}
+
+module.exports = {
+  SpeakerEmbeddingExtractor,
+  SpeakerEmbeddingManager,
+}

package/streaming-asr.js

@@ -64,5 +64,6 @@ class OnlineRecognizer {
 
 module.exports = {
   OnlineRecognizer,
+  OnlineStream,
   Display
 }