// This keeps the WebSocket connected when we are not sending any data.
private IEnumerator KeepAlive()
{
while (_listenSocket != null)
{
yield return(null);
if ((DateTime.Now - _lastKeepAlive).TotalSeconds > WsKeepAliveInterval)
{
// Temporary clip to use for KeepAlive
// TODO: Generate small sound clip to send to the service to keep alive.
//AudioClip _keepAliveClip = Resources.Load<AudioClip>("highHat");
#if ENABLE_DEBUGGING
Log.Debug("SpeechToText.KeepAlive()", "Sending keep alive.");
#endif
//_listenSocket.Send(new WSConnector.BinaryMessage(AudioClipUtil.GetL16(_keepAliveClip)));
//_keepAliveClip = null;
Debug.Log("Sending " + _prefixClips.Count + " prefix clips for keep alive");
foreach (AudioData prefixClip in _prefixClips)
{
_listenSocket.Send(new WSConnector.BinaryMessage(AudioClipUtil.GetL16(prefixClip.Clip)));
}
_prefixClips.Clear();
_lastKeepAlive = DateTime.Now;
}
}
Log.Debug("SpeechToText.KeepAlive()", "KeepAlive exited.");
}
/// <summary>
/// This function should be invoked with the AudioData input after StartListening() method has been invoked.
/// The user should continue to invoke this function until they are ready to call StopListening(), typically
/// microphone input is sent to this function.
/// </summary>
/// <param name="clip">A AudioData object containing the AudioClip and max level found in the clip.</param>
public void OnListen(AudioData clip)
{
if (m_IsListening)
{
if (m_RecordingHZ < 0)
{
m_RecordingHZ = clip.Clip.frequency;
SendStart();
}
if (!DetectSilence || clip.MaxLevel >= m_SilenceThreshold)
{
if (m_ListenActive)
{
m_ListenSocket.Send(new WSConnector.BinaryMessage(AudioClipUtil.GetL16(clip.Clip)));
m_AudioSent = true;
}
else
{
// we have not received the "listening" state yet from the server, so just queue
// the audio clips until that happens.
m_ListenRecordings.Enqueue(clip);
// check the length of this queue and do something if it gets too full.
if (m_ListenRecordings.Count > MAX_QUEUED_RECORDINGS)
{
Log.Error("SpeechToText", "Recording queue is full.");
StopListening();
if (OnError != null)
{
OnError("Recording queue is full.");
}
}
}
}
else if (m_AudioSent)
{
SendStop();
m_AudioSent = false;
}
// After sending start, we should get into the listening state within the amount of time specified
// by LISTEN_TIMEOUT. If not, then stop listening and record the error.
if (!m_ListenActive && (DateTime.Now - m_LastStartSent).TotalSeconds > LISTEN_TIMEOUT)
{
Log.Error("SpeechToText", "Failed to enter listening state.");
StopListening();
if (OnError != null)
{
OnError("Failed to enter listening state.");
}
}
}
}
private void OnListenMessage(WSConnector.Message msg)
{
if (msg is WSConnector.TextMessage)
{
WSConnector.TextMessage tm = (WSConnector.TextMessage)msg;
IDictionary json = Json.Deserialize(tm.Text) as IDictionary;
if (json != null)
{
if (json.Contains("results"))
{
SpeechResultList results = ParseRecognizeResponse(json);
if (results != null)
{
if (newSpeech == true)
{
newSpeech = false;
UnityEngine.Debug.Log("--new: " + new SpeechToTextData(results).Text);
Cloudspace.NotificationCenter.DefaultCenter().PostNotification(null, "OnListeningToUser",
new SpeechToTextData(results).Text);
}
if (finalResults == null)
{
finalResults = results;
}
else
{
SpeechResult[] aggregated = new SpeechResult[finalResults.Results.Length + results.Results.Length];
for (int i = 0; i < finalResults.Results.Length; i++)
{
aggregated [i] = finalResults.Results [i];
}
for (int i = finalResults.Results.Length; i < finalResults.Results.Length + results.Results.Length; i++)
{
aggregated [i] = results.Results [i - finalResults.Results.Length];
}
finalResults.Results = aggregated;
}
// UnityEngine.Debug.Log ("--agg: "+new SpeechToTextData (finalResults).AllText);
// when we get results, start listening for the next block ..
// if continuous is true, then we don't need to do this..
if (!EnableContinousRecognition && results.HasFinalResult())
{
SendStart();
}
// if (m_ListenCallback == null) {
StopListening();
// }
}
else
{
Log.Error("SpeechToText", "Failed to parse results: {0}", tm.Text);
}
}
else if (json.Contains("state"))
{
string state = (string)json["state"];
#if ENABLE_DEBUGGING
Log.Debug("SpeechToText", "Server state is {0}", state);
#endif
if (state == "listening")
{
if (m_IsListening)
{
if (!m_ListenActive)
{
m_ListenActive = true;
// send all pending audio clips ..
while (m_ListenRecordings.Count > 0)
{
AudioData clip = m_ListenRecordings.Dequeue();
m_ListenSocket.Send(new WSConnector.BinaryMessage(AudioClipUtil.GetL16(clip.Clip)));
m_AudioSent = true;
}
}
}
}
}
else if (json.Contains("error"))
{
string error = (string)json["error"];
Log.Error("SpeechToText", "Error: {0}", error);
StopListening();
if (OnError != null)
{
OnError(error);
}
}
else
{
Log.Warning("SpeechToText", "Unknown message: {0}", tm.Text);
}
}
else
{
Log.Error("SpeechToText", "Failed to parse JSON from server: {0}", tm.Text);
}
}
if (silenceCounter > m_PausesLimit)
{
silenceCounter = 0;
if (finalResults != null)
{
newSpeech = true;
m_ListenCallback(finalResults);
finalResults = null;
}
}
}
private void OnListenMessage(WSConnector.Message msg)
{
if (msg is WSConnector.TextMessage)
{
WSConnector.TextMessage tm = (WSConnector.TextMessage)msg;
IDictionary json = Json.Deserialize(tm.Text) as IDictionary;
if (json != null)
{
if (json.Contains("results"))
{
SpeechResultList results = ParseRecognizeResponse(json);
if (results != null)
{
// when we get results, start listening for the next block ..
// if continuous is true, then we don't need to do this..
if (!EnableContinousRecognition && results.HasFinalResult())
{
SendStart();
}
if (m_ListenCallback != null)
{
m_ListenCallback(results);
}
else
{
StopListening(); // automatically stop listening if our callback is destroyed.
}
}
else
{
Log.Error("SpeechToText", "Failed to parse results: {0}", tm.Text);
}
}
else if (json.Contains("state"))
{
string state = (string)json["state"];
#if ENABLE_DEBUGGING
Log.Debug("SpeechToText", "Server state is {0}", state);
#endif
if (state == "listening")
{
if (m_IsListening)
{
if (!m_ListenActive)
{
m_ListenActive = true;
// send all pending audio clips ..
while (m_ListenRecordings.Count > 0)
{
AudioData clip = m_ListenRecordings.Dequeue();
m_ListenSocket.Send(new WSConnector.BinaryMessage(AudioClipUtil.GetL16(clip.Clip)));
m_AudioSent = true;
}
}
}
}
}
else if (json.Contains("error"))
{
string error = (string)json["error"];
Log.Error("SpeechToText", "Error: {0}", error);
StopListening();
if (OnError != null)
{
OnError(error);
}
}
else
{
Log.Warning("SpeechToText", "Unknown message: {0}", tm.Text);
}
}
else
{
Log.Error("SpeechToText", "Failed to parse JSON from server: {0}", tm.Text);
}
}
}
private void OnListenMessage(WSConnector.Message msg)
{
if (msg is WSConnector.TextMessage)
{
WSConnector.TextMessage tm = (WSConnector.TextMessage)msg;
IDictionary json = Json.Deserialize(tm.Text) as IDictionary;
if (json != null)
{
if (json.Contains("results"))
{
SpeechRecognitionEvent results = ParseRecognizeResponse(json);
if (results != null)
{
//// when we get results, start listening for the next block ..
//if (results.HasFinalResult())
//Log.Debug("SpeechToText.OnListenMessage()", "final json response: {0}", tm.Text);
// SendStart();
if (_listenCallback != null)
{
_listenCallback(results);
}
else
{
StopListening(); // automatically stop listening if our callback is destroyed.
}
}
else
{
Log.Error("SpeechToText.OnListenMessage()", "Failed to parse results: {0}", tm.Text);
}
}
else if (json.Contains("state"))
{
string state = (string)json["state"];
#if ENABLE_DEBUGGING
Log.Debug("SpeechToText.OnListenMessage()", "Server state is {0}", state);
#endif
if (state == "listening")
{
if (_isListening)
{
if (!_listenActive)
{
_listenActive = true;
//Debug.Log("Listening, sending " + _listenRecordings.Count + " queued clips");
bool hasAudio = _listenRecordings.Count > 0;
// send all pending audio clips ..
while (_listenRecordings.Count > 0)
{
AudioData clip = _listenRecordings.Dequeue();
_listenSocket.Send(new WSConnector.BinaryMessage(AudioClipUtil.GetL16(clip.Clip)));
_audioSent = true;
}
// We may have received a stop command while waiting for the listening state.
if (_sendStopAfterListening && hasAudio)
{
SendStop();
}
}
}
}
}
else if (json.Contains("speaker_labels"))
{
SpeakerRecognitionEvent speakerRecognitionEvent = ParseSpeakerRecognitionResponse(json);
if (speakerRecognitionEvent != null)
{
_speakerLabelCallback(speakerRecognitionEvent);
}
}
else if (json.Contains("error"))
{
string error = (string)json["error"];
Log.Error("SpeechToText.OnListenMessage()", "Error: {0}", error);
StopListening();
if (OnError != null)
{
OnError(error);
}
}
else
{
Log.Warning("SpeechToText.OnListenMessage()", "Unknown message: {0}", tm.Text);
}
}
else
{
Log.Error("SpeechToText.OnListenMessage()", "Failed to parse JSON from server: {0}", tm.Text);
}
}
}
/// <summary>
/// This function should be invoked with the AudioData input after StartListening() method has been invoked.
/// The user should continue to invoke this function until they are ready to call StopListening(), typically
/// microphone input is sent to this function.
/// </summary>
/// <param name="clip">A AudioData object containing the AudioClip and max level found in the clip.</param>
/// <returns>True if audio was sent or enqueued, false if audio was discarded.</returns>
public bool OnListen(AudioData clip)
{
bool audioSentOrEnqueued = false;
_timeOffset = 0;
if (_isListening)
{
if (_recordingHZ < 0)
{
_recordingHZ = clip.Clip.frequency;
SendStart();
}
// 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 (_stopSent)
{
// Send some clips of ambient sound leading up to the audio, to improve first word recognition
if (_listenActive)
{
//Debug.Log("Sending " + _prefixClips.Count + " prefix clips");
foreach (AudioData prefixClip in _prefixClips)
{
_listenSocket.Send(new WSConnector.BinaryMessage(AudioClipUtil.GetL16(prefixClip.Clip)));
_timeOffset -= prefixClip.Clip.length;
}
}
else
{
//Debug.Log("Queuing " + _prefixClips.Count + " prefix clips");
foreach (AudioData prefixClip in _prefixClips)
{
_listenRecordings.Enqueue(clip);
_timeOffset -= prefixClip.Clip.length;
}
}
_prefixClips.Clear();
_stopSent = false;
}
if (_listenActive)
{
CurrentStatus = LexiconSpeechStatus.Sending;
_listenSocket.Send(new WSConnector.BinaryMessage(AudioClipUtil.GetL16(clip.Clip)));
_audioSent = true;
audioSentOrEnqueued = true;
}
else
{
CurrentStatus = LexiconSpeechStatus.WaitingForServer;
// we have not received the "listening" state yet from the server, so just queue
// the audio clips until that happens.
_listenRecordings.Enqueue(clip);
audioSentOrEnqueued = true;
// check the length of this queue and do something if it gets too full.
if (_listenRecordings.Count > MaxQueuedRecordings)
{
Log.Error("SpeechToText.OnListen()", "Recording queue is full.");
StopListening();
if (OnError != null)
{
OnError("Recording queue is full.");
}
}
}
}
else if (_audioSent)
{
CurrentStatus = LexiconSpeechStatus.Silence;
//Debug.Log("Send stop");
SendStop();
_audioSent = false;
_stopSent = true;
_prefixClips.Clear();
}
else
{
// Buffer some of the ambient audio for when the user starts speaking again
_prefixClips.Add(clip);
if (_prefixClips.Count > _prefixClipCount)
{
_prefixClips.RemoveAt(0);
}
}
// After sending start, we should get into the listening state within the amount of time specified
// by LISTEN_TIMEOUT. If not, then stop listening and record the error.
if (!_listenActive && (DateTime.Now - _lastStartSent).TotalSeconds > ListenTimeout)
{
Log.Error("SpeechToText.OnListen()", "Failed to enter listening state.");
StopListening();
if (OnError != null)
{
OnError("Failed to enter listening state.");
}
}
}
return(audioSentOrEnqueued);
}
public override byte[] ToBinary()
{
return(AudioClipUtil.GetL16(m_Audio.Clip));
}