Merge pull request #296 from kimberlysiva/279-detectSilence

Fix detect silence to work with smaller audio chunks

Author: mediumTaj

Examples/ServiceExamples/Scripts/ExampleStreaming.cs

@@ -63,10 +63,10 @@ public bool Active
         {
             if (value && !_speechToText.IsListening)
             {
-                _speechToText.DetectSilence = false;
+                _speechToText.DetectSilence = true;
                 _speechToText.EnableWordConfidence = true;
                 _speechToText.EnableTimestamps = true;
-                _speechToText.SilenceThreshold = 0.1f;
+                _speechToText.SilenceThreshold = 0.01f;
                 _speechToText.MaxAlternatives = 0;
                 _speechToText.EnableInterimResults = true;
                 _speechToText.OnError = OnError;
@@ -177,7 +177,7 @@ private IEnumerator RecordingHandler()
 
                 //  Create AudioData and use the samples we just created
                 AudioData record = new AudioData();
-                record.MaxLevel = Mathf.Max(samples);
+                record.MaxLevel = Mathf.Max(Mathf.Abs(Mathf.Min(samples)), Mathf.Max(samples));
                 record.Clip = AudioClip.Create("Recording", chunkSize, _recording.channels, _recordingHZ, false);
                 record.Clip.SetData(samples, 0);
 

Scripts/Services/SpeechToText/v1/SpeechToText.cs

@@ -109,6 +109,8 @@ public class SpeechToText : IWatsonService
         private string _acoustic_customization_id = null;
         private float _customization_weight = 0.3f;
         private bool _streamMultipart = false;           //  If true sets `Transfer-Encoding` header of multipart request to `chunked`.
+        private float _silenceDuration = 0.0f;
+        private float _silenceCutoff = 1.0f;
 
         private fsSerializer _serializer = new fsSerializer();
         private Credentials _credentials = null;
@@ -465,7 +467,17 @@ public void OnListen(AudioData clip)
                     SendStart();
                 }
 
-                if (!DetectSilence || clip.MaxLevel >= _silenceThreshold)
+                // If silence persists for _silenceCutoff seconds, send stop and discard clips until audio resumes
+                if (DetectSilence && clip.MaxLevel < _silenceThreshold)
+                {
+                    _silenceDuration += clip.Clip.length;
+                }
+                else
+                {
+                    _silenceDuration = 0.0f;
+                }
+
+                if (!DetectSilence || _silenceDuration < _silenceCutoff)
                 {
                     if (_listenActive)
                     {