Add VAD platform capability through webrtcvad

package.json

@@ -23,12 +23,13 @@
     "byline": "^5.0.0",
     "canberra": "^0.1.2",
     "dbus-native": "^0.4.0",
-    "genie-toolkit": "~0.8.0",
+    "genie-toolkit": "github:stanford-oval/genie-toolkit",
     "keytar": "^7.7.0",
     "node-gettext": "^3.0.0",
     "pulseaudio2": "^0.5.0",
     "snowboy": "^1.3.1",
-    "thingpedia": "~2.9.0"
+    "thingpedia": "~2.9.0",
+    "webrtcvad": "^1.0.1"
   },
   "devDependencies": {
     "eslint": "^7.27.0"

service/config.js

@@ -21,5 +21,4 @@
 
 module.exports.SEMPRE_URL = 'https://nlp.almond.stanford.edu';
 module.exports.THINGPEDIA_URL = 'https://thingpedia.stanford.edu/thingpedia';
-module.exports.MS_SPEECH_RECOGNITION_PRIMARY_KEY = 'de1f02817356494483ba502b2ce95f6f';
-module.exports.MS_SPEECH_RECOGNITION_SECONDARY_KEY = '3dc6ce0b832940f0b0c984a1517c457e';
+module.exports.NL_URL = process.env.THINGENGINE_NL_URL || 'https://nlp-staging.almond.stanford.edu';

service/main.js

@@ -338,7 +338,7 @@ class AppControlChannel extends events.EventEmitter {
         this._history = [];
 
         this._speechHandler = new Genie.SpeechHandler(this._conversation, _engine.platform, {
-            subscriptionKey: Config.MS_SPEECH_RECOGNITION_PRIMARY_KEY
+            nlUrl: Config.NL_URL
         });
 
         let play;

service/platform/index.js

@@ -258,6 +258,45 @@ class SystemSettings {
     }
 }
 
+let webrtcvad;
+try {
+    webrtcvad = require('webrtcvad').default;
+} catch(e) {
+    console.log("VAD not available");
+    webrtcvad = null;
+}
+
+class VAD {
+    constructor() {
+        this._instance = null;
+        this.frameSize = 0;
+    }
+
+    setup(bitrate, level) {
+        if (this._instance)
+            this._instance = null;
+
+        if (webrtcvad) {
+            this._instance = new webrtcvad(bitrate, level);
+            // 16khz audio single-channel 16 bit: 10ms: 160b, 20ms: 320b, 30ms: 480b
+            this.frameSize = 320;
+            // console.log("setup VAD bitrate", bitrate, "level", level);
+            return true;
+        }
+
+        return false;
+    }
+
+    process(chunk) {
+        if (!this._instance)
+            return false;
+        let n = chunk.length % this.frameSize, r = 0;
+        for (let i = 0; i < n; i++)
+            r += this._instance.process(chunk.slice(i * this.frameSize, this.frameSize));
+        return r;
+    }
+}
+
 class Platform extends Tp.BasePlatform {
     // Initialize the platform code
     // Will be called before instantiating the engine
@@ -300,6 +339,9 @@ class Platform extends Tp.BasePlatform {
             this._ensurePulseConfig();
         });
         this._wakeWordDetector = new WakeWordDetector();
+        this._voiceDetector = null;
+        if (webrtcvad && VAD)
+            this._voiceDetector = new VAD();
 
         this._sqliteKey = null;
     }
@@ -428,6 +470,9 @@ class Platform extends Tp.BasePlatform {
         case 'pulseaudio':
             return true;
 
+        case 'voice-detector':
+            return this._voiceDetector !== null;
+
         case 'bluetooth':
             // temporarily disabled
             return false;
@@ -468,7 +513,8 @@ class Platform extends Tp.BasePlatform {
 
         case 'wakeword-detector':
             return this._wakeWordDetector;
-
+        case 'voice-detector':
+            return this._voiceDetector;
         case 'app-launcher':
             return this._appLauncher;
         case 'system-lock':