add support for vad to gather and transcribe (#67)

lib/tasks/gather.js

@@ -30,6 +30,10 @@ class TaskGather extends Task {
       this.hints = recognizer.hints || [];
       this.altLanguages = recognizer.altLanguages || [];
 
+      /* vad: if provided, we dont connect to recognizer until voice activity is detected */
+      const {enable, voiceMs = 0, mode = -1} = recognizer.vad || {};
+      this.vad = {enable, voiceMs, mode};
+
       /* aws options */
       this.vocabularyName = recognizer.vocabularyName;
       this.vocabularyFilterName = recognizer.vocabularyFilterName;
@@ -137,6 +141,12 @@ class TaskGather extends Task {
   async _initSpeech(cs, ep) {
     const opts = {};
 
+    if (this.vad.enable) {
+      opts.START_RECOGNIZING_ON_VAD = 1;
+      if (this.vad.voiceMs) opts.RECOGNIZER_VAD_VOICE_MS = this.vad.voiceMs;
+      if (this.vad.mode >= 0 && this.vad.mode <= 3) opts.RECOGNIZER_VAD_MODE = this.vad.mode;
+    }
+
     if ('google' === this.vendor) {
       if (this.sttCredentials) opts.GOOGLE_APPLICATION_CREDENTIALS = JSON.stringify(this.sttCredentials.credentials);
       Object.assign(opts, {

lib/tasks/specs.json

@@ -389,6 +389,7 @@
         "enum": ["google", "aws", "microsoft", "default"]
       },
       "language": "string",
+      "vad": "#vad",
       "hints": "array",
       "altLanguages": "array",
       "profanityFilter": "boolean",
@@ -457,5 +458,15 @@
     "required": [
       "name"
     ]
+  },
+  "vad": {
+    "properties": {
+      "enable": "boolean",
+      "voiceMs": "number",
+      "mode": "number"  
+    },
+    "required": [
+      "enable"
+    ]
   }
 }

lib/tasks/transcribe.js

@@ -22,6 +22,10 @@ class TaskTranscribe extends Task {
     this.interim = !!recognizer.interim;
     this.separateRecognitionPerChannel = recognizer.separateRecognitionPerChannel;
 
+    /* vad: if provided, we dont connect to recognizer until voice activity is detected */
+    const {enable, voiceMs = 0, mode = -1} = recognizer.vad || {};
+    this.vad = {enable, voiceMs, mode};
+
     /* google-specific options */
     this.hints = recognizer.hints || [];
     this.profanityFilter = recognizer.profanityFilter;
@@ -105,6 +109,12 @@ class TaskTranscribe extends Task {
   async _startTranscribing(cs, ep) {
     const opts = {};
 
+    if (this.vad.enable) {
+      opts.START_RECOGNIZING_ON_VAD = 1;
+      if (this.vad.voiceMs) opts.RECOGNIZER_VAD_VOICE_MS = this.vad.voiceMs;
+      if (this.vad.mode >= 0 && this.vad.mode <= 3) opts.RECOGNIZER_VAD_MODE = this.vad.mode;
+    }
+
     ep.addCustomEventListener(GoogleTranscriptionEvents.Transcription, this._onTranscription.bind(this, cs, ep));
     ep.addCustomEventListener(GoogleTranscriptionEvents.NoAudioDetected, this._onNoAudio.bind(this, cs, ep));
     ep.addCustomEventListener(GoogleTranscriptionEvents.MaxDurationExceeded,