/// <summary>
///
/// </summary>
/// <param name="encoding"></param>
/// <param name="sampleRateHertz"></param>
/// <param name="languageCode"></param>
/// <param name="intermResults"></param>
public async Task ConfigureSpeechRequest(
RecognitionConfig.Types.AudioEncoding encoding,
int sampleRateHertz,
string languageCode,
bool intermResults)
{
if (null == speechClient)
{
return;
}
streamingCall = speechClient.StreamingRecognize();
request = new StreamingRecognizeRequest()
{
StreamingConfig = new StreamingRecognitionConfig()
{
Config = new RecognitionConfig()
{
Encoding = encoding,
SampleRateHertz = sampleRateHertz,
LanguageCode = languageCode,
},
InterimResults = intermResults,
}
};
// Write the initial request with the config.
await streamingCall.WriteAsync(request);
}
private async Task StartStreamAsync()
{
var now = DateTime.UtcNow;
if (m_rpcStream != null && now >= m_rpcStreamDeadline)
{
Console.WriteLine($"Closing stream before it times out");
await m_rpcStream.WriteCompleteAsync();
m_rpcStream.GrpcCall.Dispose();
m_rpcStream = null;
}
// If we have a valid stream at this point, we're fine.
if (m_rpcStream != null)
{
return;
}
// We need to create a new stream, either because we're just starting or because we've just closed the previous one.
m_rpcStream = m_client.StreamingRecognize();
m_rpcStreamDeadline = now + s_streamTimeLimit;
m_processingBufferStart = TimeSpan.Zero;
m_serverResponseAvailableTask = m_rpcStream.GetResponseStream().MoveNextAsync();
await m_rpcStream.WriteAsync(m_streamingRecognizeRequest);
Console.WriteLine($"Writing {m_speechUnprocessedBuffer.Count} chunks into the new stream.");
foreach (var chunk in m_speechUnprocessedBuffer)
{
await WriteAudioChunk(chunk);
}
}
/// <summary>
/// Takes a single sample chunk from the microphone buffer, keeps a local copy
/// (in case we need to send it again in a new request) and sends it to the server.
/// </summary>
/// <returns></returns>
private async Task TransferMicrophoneChunkAsync()
{
// This will block - but only for ~100ms, unless something's really broken.
var chunk = _microphoneBuffer.Take();
//Console.WriteLine(("ProcessingBuffer AddLast" + chunk.Length.ToString()));
_processingBuffer.AddLast(chunk);
try
{
await WriteAudioChunk(chunk);
if (_hasText && (DateTime.UtcNow - _lastText).TotalSeconds > 2)
{
Console.WriteLine("pause");
_processingBuffer.Clear();
this._rpcStreamDeadline = DateTime.UtcNow.AddSeconds(-1);
await resendLastTextToBridgeAsFinalAsync();
_hasText = false;
}
}
catch (Grpc.Core.RpcException e)
{
Console.WriteLine("Exception writing to google - timeout\n{0}", e.Message);
_rpcStream = null;
}
}
/// <summary>
/// Starts a new RPC streaming call if necessary. This will be if either it's the first call
/// (so we don't have a current request) or if the current request will time out soon.
/// In the latter case, after starting the new request, we copy any chunks we'd already sent
/// in the previous request which hadn't been included in a "final result".
/// </summary>
private async Task MaybeStartStreamAsync()
{
var now = DateTime.UtcNow;
if (_rpcStream != null && now >= _rpcStreamDeadline)
{
Console.WriteLine($"Closing stream before it times out");
await _rpcStream.WriteCompleteAsync();
_rpcStream.GrpcCall.Dispose();
_rpcStream = null;
}
// If we have a valid stream at this point, we're fine.
if (_rpcStream != null)
{
//Console.WriteLine("We already have a google stream");
return;
}
var translationLanguage = MapLanguageCodeToGoogleLanguage(_myct.FromLang);
Console.WriteLine($"Creating new google stream to translate from {translationLanguage}");
// We need to create a new stream, either because we're just starting or because we've just closed the previous one.
_rpcStream = _client.StreamingRecognize();
_rpcStreamDeadline = now + s_streamTimeLimit;
_processingBufferStart = TimeSpan.Zero;
_serverResponseAvailableTask = _rpcStream.GetResponseStream().MoveNextAsync();
await _rpcStream.WriteAsync(new StreamingRecognizeRequest
{
StreamingConfig = new StreamingRecognitionConfig
{
Config = new RecognitionConfig
{
Encoding = RecognitionConfig.Types.AudioEncoding.Linear16,
SampleRateHertz = SampleRate,
LanguageCode = translationLanguage,// _myct.FromLang,//"en-US",
MaxAlternatives = 1,
UseEnhanced = true,
EnableAutomaticPunctuation = true
},
InterimResults = true,
//SingleUtterance=false
}
});
Console.WriteLine($"Writing {_processingBuffer.Count} chunks into the new stream.");
foreach (var chunk in _processingBuffer)
{
await WriteAudioChunk(chunk);
}
}
public async Task Start(string lang)
{
var speech = SpeechClient.Create(_channel);
WriteMore = true;
lock (WriteLock) WriteMore = true;
StreamingCall = speech.StreamingRecognize();
await StreamingCall.WriteAsync(ConfigRequest(lang));
WaveIn.StartRecording();
Console.WriteLine("Speak now.");
}
/// <summary>Snippet for StreamingRecognize</summary>
public async Task StreamingRecognize()
{
// Snippet: StreamingRecognize(CallSettings, BidirectionalStreamingSettings)
// Create client
SpeechClient speechClient = SpeechClient.Create();
// Initialize streaming call, retrieving the stream object
SpeechClient.StreamingRecognizeStream response = speechClient.StreamingRecognize();
// Sending requests and retrieving responses can be arbitrarily interleaved
// Exact sequence will depend on client/server behavior
// Create task to do something with responses from server
Task responseHandlerTask = Task.Run(async() =>
{
// Note that C# 8 code can use await foreach
AsyncResponseStream <StreamingRecognizeResponse> responseStream = response.GetResponseStream();
while (await responseStream.MoveNextAsync())
{
StreamingRecognizeResponse responseItem = responseStream.Current;
// Do something with streamed response
}
// The response stream has completed
});
// Send requests to the server
bool done = false;
while (!done)
{
// Initialize a request
StreamingRecognizeRequest request = new StreamingRecognizeRequest
{
StreamingConfig = new StreamingRecognitionConfig(),
AudioContent = ByteString.Empty,
};
// Stream a request to the server
await response.WriteAsync(request);
// Set "done" to true when sending requests is complete
}
// Complete writing requests to the stream
await response.WriteCompleteAsync();
// Await the response handler
// This will complete once all server responses have been processed
await responseHandlerTask;
// End snippet
}
/// <summary>
/// Starts a new RPC streaming call if necessary. This will be if either it's the first call
/// (so we don't have a current request) or if the current request will time out soon.
/// In the latter case, after starting the new request, we copy any chunks we'd already sent
/// in the previous request which hadn't been included in a "final result".
/// </summary>
private async Task MaybeStartStreamAsync()
{
var now = DateTime.UtcNow;
if (_rpcStream != null && now >= _rpcStreamDeadline)
{
Console.WriteLine($"Closing stream before it times out");
await _rpcStream.WriteCompleteAsync();
_rpcStream.GrpcCall.Dispose();
_rpcStream = null;
}
// If we have a valid stream at this point, we're fine.
if (_rpcStream != null)
{
return;
}
// We need to create a new stream, either because we're just starting or because we've just closed the previous one.
_rpcStream = _client.StreamingRecognize();
_rpcStreamDeadline = now + s_streamTimeLimit;
_processingBufferStart = TimeSpan.Zero;
_serverResponseAvailableTask = _rpcStream.GetResponseStream().MoveNextAsync();
await _rpcStream.WriteAsync(new StreamingRecognizeRequest
{
StreamingConfig = new StreamingRecognitionConfig
{
Config = new RecognitionConfig
{
Encoding = RecognitionConfig.Types.AudioEncoding.Linear16,
SampleRateHertz = SampleRate,
LanguageCode = "en-US",
MaxAlternatives = 1
},
InterimResults = true,
}
});
Console.WriteLine($"Writing {_processingBuffer.Count} chunks into the new stream.");
foreach (var chunk in _processingBuffer)
{
await WriteAudioChunk(chunk);
}
}
/// <summary>Snippet for StreamingRecognize</summary>
public async Task StreamingRecognize()
{
// Snippet: StreamingRecognize(CallSettings,BidirectionalStreamingSettings)
// Create client
SpeechClient speechClient = SpeechClient.Create();
// Initialize streaming call, retrieving the stream object
SpeechClient.StreamingRecognizeStream duplexStream = speechClient.StreamingRecognize();
// Sending requests and retrieving responses can be arbitrarily interleaved.
// Exact sequence will depend on client/server behavior.
// Create task to do something with responses from server
Task responseHandlerTask = Task.Run(async() =>
{
IAsyncEnumerator <StreamingRecognizeResponse> responseStream = duplexStream.ResponseStream;
while (await responseStream.MoveNext())
{
StreamingRecognizeResponse response = responseStream.Current;
// Do something with streamed response
}
// The response stream has completed
});
// Send requests to the server
bool done = false;
while (!done)
{
// Initialize a request
StreamingRecognizeRequest request = new StreamingRecognizeRequest();
// Stream a request to the server
await duplexStream.WriteAsync(request);
// Set "done" to true when sending requests is complete
}
// Complete writing requests to the stream
await duplexStream.WriteCompleteAsync();
// Await the response handler.
// This will complete once all server responses have been processed.
await responseHandlerTask;
// End snippet
}
public GoogleSpeechSession(string socketId, GoogleSessionConfig config, Action <string, string[]> processTranscripts)
{
try
{
_log.Debug($"Creating google api session: {socketId} - {config}");
var speech = SpeechClient.Create();
_streamingCall = speech.StreamingRecognize();
SockedId = socketId;
Config = config;
ProcessTranscripts = processTranscripts;
_audioFileName = $"{DateTime.Now:yyyyMMddHHmmss}.wav";
_audioFileStream = File.OpenWrite(Path.Combine($@"c:\temp", _audioFileName));
}
catch (Exception e)
{
_log.Error(e);
}
}
private async Task SetupSpeechClient()
{
_speech = SpeechClient.Create();
_streamingCall = _speech.StreamingRecognize();
await _streamingCall.WriteAsync(
new StreamingRecognizeRequest()
{
StreamingConfig = new StreamingRecognitionConfig()
{
Config = new RecognitionConfig()
{
Encoding = RecognitionConfig.Types.AudioEncoding.Linear16,
SampleRateHertz = 16000,
LanguageCode = "sv",
},
InterimResults = true,
}
});
}
/// <summary>
/// Check if current prc stream exceed its life. If so end current one and start a new one.
/// </summary>
private async Task CreateNewStreamIfNeeded()
{
if (_recognizeStream == null) // The first one
{
lock (_recognizeStreamLock)
{
_rpcStreamDeadline = DateTime.UtcNow + TimeSpan.FromSeconds(RpcStreamLife);
_recognizeStream = GoogleCloudSpeechClient.StreamingRecognize();
}
await _recognizeStream.WriteAsync(new StreamingRecognizeRequest()
{
StreamingConfig = _recognitionConfig
}); //Configured
_wasapiCapture.DataAvailable += OnWaveInDataAvailable;
}
else if (DateTime.UtcNow >= _rpcStreamDeadline) // Expiring, switch to new
{
_wasapiCapture.DataAvailable -= OnWaveInDataAvailable; // Stop sending new bytes
SpeechClient.StreamingRecognizeStream oldStream;
lock (_recognizeStreamLock)
{
oldStream = _recognizeStream;
_rpcStreamDeadline = DateTime.UtcNow + TimeSpan.FromSeconds(RpcStreamLife);
_recognizeStream = GoogleCloudSpeechClient.StreamingRecognize(); // Create new one
}
await _recognizeStream.WriteAsync(new StreamingRecognizeRequest()
{
StreamingConfig = _recognitionConfig
}); //Configure new one
_wasapiCapture.DataAvailable += OnWaveInDataAvailable; // Start sending to new stream
await oldStream.WriteCompleteAsync(); // Complete old one
oldStream.GrpcCall.Dispose();
}
}
public async void Start()
{
try
{
var credentials = GoogleCredential.FromFile("Extensions/25670f9058d6.json");
var channel = new Channel(SpeechClient.DefaultEndpoint.Host, credentials.ToChannelCredentials());
timer = new Timer(Elapsed, null, 60000, Timeout.Infinite);
speech = SpeechClient.Create(channel);
stream = new SpeechStream(65536);
streaming = speech.StreamingRecognize();
engine = new SpeechRecognitionEngine();
engine.LoadGrammar(new DictationGrammar());
engine.AudioStateChanged += Engine_AudioStateChanged;
engine.SetInputToAudioStream(stream, new SpeechAudioFormatInfo(EncodingFormat.ULaw, 8000, 8, 1, 8000, 1, null));
engine.RecognizeAsync(RecognizeMode.Multiple);
}
catch (Exception ex)
{
systemLog.Error(prefix + " Could not create SpeechClient", ex);
}
await streaming.WriteAsync(new StreamingRecognizeRequest()
{
StreamingConfig = new StreamingRecognitionConfig()
{
InterimResults = true,
Config = new RecognitionConfig()
{
Encoding = RecognitionConfig.Types.AudioEncoding.Mulaw,
SampleRateHertz = 8000,
LanguageCode = "en-US"
},
}
});
systemLog.DebugFormat("{0} Google Speech To Text stream started", prefix);
IsReady = true;
handle = Task.Run(async() =>
{
while (await streaming.ResponseStream.MoveNext(default(CancellationToken)))
{
try
{
foreach (var result in streaming.ResponseStream.Current.Results)
{
if (!result.IsFinal) continue;
if (result.Alternatives.Count == 0) continue;
var alternative = result.Alternatives.OrderByDescending(a => a.Confidence).FirstOrDefault();
var args = new TextEventArgs(alternative.Transcript, Encoding.UTF8);
TextAvailable(this, args);
}
}
catch
{
break;
}
}
});
}
/** (Re-)initializes the Cloud-based streaming speech recognizer. */
private void ReInitStreamRecognizer()
{
lock (speakerIdBufferLock)
{
speakerIdBufferPos = 0;
}
recogStream = speechClient.StreamingRecognize();
SpeakerDiarizationConfig diarizationConfig = new SpeakerDiarizationConfig()
{
EnableSpeakerDiarization = ENABLE_SPEAKER_DIARIZATION,
MaxSpeakerCount = MAX_SPEAKER_COUNT,
MinSpeakerCount = MIN_SPEAKER_COUNT,
};
recogStream.WriteAsync(new StreamingRecognizeRequest()
{
StreamingConfig = new StreamingRecognitionConfig()
{
Config = new RecognitionConfig()
{
Encoding = RecognitionConfig.Types.AudioEncoding.Linear16,
AudioChannelCount = 1,
SampleRateHertz = audioFormat.SampleRate,
LanguageCode = LANGUAGE_CODE,
DiarizationConfig = diarizationConfig,
},
SingleUtterance = false,
},
});;
Task.Run(async() =>
{
while (await recogStream.GetResponseStream().MoveNextAsync())
{
foreach (var result in recogStream.GetResponseStream().Current.Results)
{
if (result.Alternatives.Count == 0)
{
continue;
}
// Identify the alternative with the highest confidence.
SpeechRecognitionAlternative bestAlt = null;
foreach (var alternative in result.Alternatives)
{
if (bestAlt == null || alternative.Confidence > bestAlt.Confidence)
{
bestAlt = alternative;
}
}
string transcript = bestAlt.Transcript.Trim();
if (transcript.Length == 0)
{
continue;
}
string transcriptInfo =
$"Speech transcript: {DateTime.Now}: \"" +
$"{transcript}\" (confidence={bestAlt.Confidence})";
if (ENABLE_SPEAKER_DIARIZATION)
{
int speakerTag = bestAlt.Words[bestAlt.Words.Count - 1].SpeakerTag;
transcriptInfo += $" (speakerTag={speakerTag})";
}
Debug.WriteLine(transcriptInfo);
if (ENABLE_SPEAKER_DIARIZATION && ENABLE_SPEAKER_ID)
{
recognizeSpeaker(transcript, bestAlt);
}
}
}
});
cummulativeRecogSeconds = 0f;
}
public AudioAmplitudeStream(SpeechClient.StreamingRecognizeStream streamingCall)
{
writeLock = new object();
_streamingCall = streamingCall;
}
private static async Task <object> StreamingMicRecognizeAsync(int seconds)
{
if (WaveIn.DeviceCount < 1)
{
File.WriteAllText("error.txt", "No microphone!");
return((object)-1);
}
string lower = INISetting.GetValueWithAdd <string>("CredentialsFilePath", "credentials.json").ToLower();
Console.WriteLine(lower);
GoogleCredential googleCredential;
using (Stream stream = (Stream) new FileStream(lower, FileMode.Open))
googleCredential = GoogleCredential.FromStream(stream);
SpeechClient.StreamingRecognizeStream streamingCall = SpeechClient.Create(new Channel(SpeechClient.DefaultEndpoint.Host, googleCredential.ToChannelCredentials())).StreamingRecognize();
await streamingCall.WriteAsync(new StreamingRecognizeRequest()
{
StreamingConfig = new StreamingRecognitionConfig()
{
Config = new RecognitionConfig()
{
Encoding = RecognitionConfig.Types.AudioEncoding.Linear16,
SampleRateHertz = 16000,
LanguageCode = "ru"
},
InterimResults = true
}
});
Task printResponses = Task.Run((Func <Task>)(async() =>
{
string s = "";
while (true)
{
if (await streamingCall.ResponseStream.MoveNext(new CancellationToken()))
{
using (IEnumerator <StreamingRecognitionResult> enumerator1 = streamingCall.ResponseStream.Current.Results.GetEnumerator())
{
if (enumerator1.MoveNext())
{
using (IEnumerator <SpeechRecognitionAlternative> enumerator2 = enumerator1.Current.Alternatives.GetEnumerator())
{
if (enumerator2.MoveNext())
{
SpeechRecognitionAlternative current = enumerator2.Current;
Console.WriteLine(current.Transcript);
s += current.Transcript;
}
}
}
}
File.WriteAllText(Path.GetTempPath() + "\\speechtext\\speechtext.txt", s);
s = "";
}
else
{
break;
}
}
}));
object writeLock = new object();
bool writeMore = true;
WaveInEvent waveIn = new WaveInEvent();
waveIn.DeviceNumber = 0;
waveIn.WaveFormat = new WaveFormat(16000, 1);
waveIn.DataAvailable += (EventHandler <WaveInEventArgs>)((sender, args) =>
{
lock (writeLock)
{
if (!writeMore)
{
return;
}
streamingCall.WriteAsync(new StreamingRecognizeRequest()
{
AudioContent = ByteString.CopyFrom(args.Buffer, 0, args.BytesRecorded)
}).Wait();
}
});
waveIn.StartRecording();
Console.WriteLine("Speak now " + (object)seconds);
await Task.Delay(TimeSpan.FromSeconds((double)seconds));
waveIn.StopRecording();
lock (writeLock)
writeMore = false;
await streamingCall.WriteCompleteAsync();
await printResponses;
return((object)0);
}
