azure.cognitiveservices.speech SpeechRecognizer canceled automatically

Question

azure.cognitiveservices.speech SpeechRecognizer canceled automatically

Mia17 41

I am using azure.cognitiveservices.speech Service to convert an wav content into text

Below is the message I got.

It seems to be canceled by some reason and didn't generate any sentences.

But I check the wav file is good, don't know how to fix it.

SpeechRecognitionCanceledEventArgs(session_id=fbae997c309d4f5bae4867cf1ab7f508, result=SpeechRecognitionResult(result_id=ddf647f11ae643d0a2c212473ac358d0, text="", reason=ResultReason.Canceled))

SessionEventArgs(session_id=fbae997c309d4f5bae4867cf1ab7f508)

Here's my code

romungi-MSFT 30,291 Reputation points Microsoft Employee

2022-08-25T12:13:34.69+00:00
@Mia17 Could you add logging to your SDK to check if you see more details for the cancellation?

speech_config.set_property(speechsdk.PropertyId.Speech_LogFilename, "LogfilePathAndName")

The cancellation could be for multiple reasons platform, file format or incorrect speech resource configuration. Was any of your previous files or mic to recognize speech before with the SDK that worked with similar method?
Mia17 41 Reputation points

2022-08-26T02:11:27.537+00:00

Here's my log file

One row shows that message
Does it mean the audio file is too big?

[97991]: 47569ms SPX_DBG_TRACE_VERBOSE: named_properties.h:281 ISpxPropertyBagImpl::SetStringValue: this=0x0x7fd6a806bda0; name='RESULT-ErrorDetails'; value='Due to service inactivity, the client buffer exceeded maximum size. Resetting the buffer.'

234949-speech-to-txt.log

romungi-MSFT 30,291 Microsoft Employee

@Mia17 Could you add additional callbacks similar to this snippet and run the method again?

 # Connect callbacks to the events fired by the speech recognizer  
    speech_recognizer.recognizing.connect(lambda evt: print('RECOGNIZING: {}'.format(evt)))  
    speech_recognizer.recognized.connect(lambda evt: print('RECOGNIZED: {}'.format(evt)))  
    speech_recognizer.session_started.connect(lambda evt: print('SESSION STARTED: {}'.format(evt)))  
    speech_recognizer.session_stopped.connect(lambda evt: print('SESSION STOPPED {}'.format(evt)))  
    speech_recognizer.canceled.connect(lambda evt: print('CANCELED {}'.format(evt)))  
    # stop continuous recognition on either session stopped or canceled events  
    speech_recognizer.session_stopped.connect(stop_cb)  
    speech_recognizer.canceled.connect(stop_cb)  
  
    # Start continuous speech recognition  
    speech_recognizer.start_continuous_recognition()  
    while not done:  
        time.sleep(.5)  
  
    speech_recognizer.stop_continuous_recognition()

If the audio file is big you could try with a sample file, if the setup isn't correct it should fail with the same error.

Mia17 41 Reputation points

2022-08-30T04:34:04.187+00:00
I tried 4 different wav files with the same code, but only 1 file worked

Each wav file looks fine when I played them

The worked wav file is bigger and longer than the failed wav , so I think it's not the size issue

The successed wav file：125MB、duration 1:08:31
Thi failed wav file：105MB、duration 57:19

Below are 3 log files

The sample file wav log (success)

The successed wav log (remove confidential log)

The failed wav log

235963-whatstheweatherlike.log

235939-speech-to-txt-successed.log

235954-speech-to-txt-failed.log

romungi-MSFT 30,291 Microsoft Employee

@Mia17 Thanks for sharing. I think the problem is most probably something to do with the content of the failed file, but I am not sure what could have caused the SDK to not process the same.

    [133546]: 47693ms SPX_DBG_TRACE_VERBOSE:  audio_stream_session.cpp:1186 [0x5600551cc730]CSpxAudioStreamSession::StopRecognizing: We've been asked to stop whatever it is we're doing, while we're actively processing audio ...  
    [78809]: 47693ms SPX_DBG_TRACE_VERBOSE:  handle_table.h:104 CSpxHandleTable::TrackHandle t=N9Microsoft17CognitiveServices6Speech4Impl21ISpxRecognitionResultE h=0x0x7ff7c006bd70, p=0x0x7ff7c006bd70, tot=1  
    [133546]: 47693ms SPX_DBG_TRACE_VERBOSE:  audio_stream_session.cpp:1189   
[name='RESULT-ErrorDetails'; value='Due to service inactivity, the client buffer exceeded maximum size. Resetting the buffer. SessionId: f63594b7822b42f282574faf38b0eee4'  
    [133546]: 47693ms SPX_DBG_TRACE_VERBOSE:  audio_pump.cpp:107 [0x5600551c8820]CSpxAudioPump::StopPump() ... mutex LOCKED

Could you please add an issue to the SDK repo with the above logs? This would help the SDK team to investigate and advise what could be wrong with this session. Thanks!!

romungi-MSFT 30,291 Reputation points Microsoft Employee

2022-08-25T12:13:34.69+00:00

@Mia17 Could you add logging to your SDK to check if you see more details for the cancellation?

speech_config.set_property(speechsdk.PropertyId.Speech_LogFilename, "LogfilePathAndName")

The cancellation could be for multiple reasons platform, file format or incorrect speech resource configuration. Was any of your previous files or mic to recognize speech before with the SDK that worked with similar method?
Mia17 41 Reputation points

2022-08-26T02:11:27.537+00:00

Here's my log file

One row shows that message
Does it mean the audio file is too big?

[97991]: 47569ms SPX_DBG_TRACE_VERBOSE: named_properties.h:281 ISpxPropertyBagImpl::SetStringValue: this=0x0x7fd6a806bda0; name='RESULT-ErrorDetails'; value='Due to service inactivity, the client buffer exceeded maximum size. Resetting the buffer.'

234949-speech-to-txt.log
romungi-MSFT 30,291 Reputation points Microsoft Employee

2022-08-29T07:40:37.027+00:00

@Mia17 Could you add additional callbacks similar to this snippet and run the method again?

# Connect callbacks to the events fired by the speech recognizer speech_recognizer.recognizing.connect(lambda evt: print('RECOGNIZING: {}'.format(evt))) speech_recognizer.recognized.connect(lambda evt: print('RECOGNIZED: {}'.format(evt))) speech_recognizer.session_started.connect(lambda evt: print('SESSION STARTED: {}'.format(evt))) speech_recognizer.session_stopped.connect(lambda evt: print('SESSION STOPPED {}'.format(evt))) speech_recognizer.canceled.connect(lambda evt: print('CANCELED {}'.format(evt))) # stop continuous recognition on either session stopped or canceled events speech_recognizer.session_stopped.connect(stop_cb) speech_recognizer.canceled.connect(stop_cb) # Start continuous speech recognition speech_recognizer.start_continuous_recognition() while not done: time.sleep(.5) speech_recognizer.stop_continuous_recognition()

If the audio file is big you could try with a sample file, if the setup isn't correct it should fail with the same error.
Mia17 41 Reputation points

2022-08-30T04:34:04.187+00:00

I tried 4 different wav files with the same code, but only 1 file worked

Each wav file looks fine when I played them

The worked wav file is bigger and longer than the failed wav , so I think it's not the size issue

The successed wav file：125MB、duration 1:08:31
Thi failed wav file：105MB、duration 57:19

Below are 3 log files

The sample file wav log (success)

The successed wav log (remove confidential log)

The failed wav log

235963-whatstheweatherlike.log

235939-speech-to-txt-successed.log

235954-speech-to-txt-failed.log
romungi-MSFT 30,291 Reputation points Microsoft Employee

2022-08-30T12:26:09.053+00:00

@Mia17 Thanks for sharing. I think the problem is most probably something to do with the content of the failed file, but I am not sure what could have caused the SDK to not process the same.

[133546]: 47693ms SPX_DBG_TRACE_VERBOSE: audio_stream_session.cpp:1186 [0x5600551cc730]CSpxAudioStreamSession::StopRecognizing: We've been asked to stop whatever it is we're doing, while we're actively processing audio ... [78809]: 47693ms SPX_DBG_TRACE_VERBOSE: handle_table.h:104 CSpxHandleTable::TrackHandle t=N9Microsoft17CognitiveServices6Speech4Impl21ISpxRecognitionResultE h=0x0x7ff7c006bd70, p=0x0x7ff7c006bd70, tot=1 [133546]: 47693ms SPX_DBG_TRACE_VERBOSE: audio_stream_session.cpp:1189 [name='RESULT-ErrorDetails'; value='Due to service inactivity, the client buffer exceeded maximum size. Resetting the buffer. SessionId: f63594b7822b42f282574faf38b0eee4' [133546]: 47693ms SPX_DBG_TRACE_VERBOSE: audio_pump.cpp:107 [0x5600551c8820]CSpxAudioPump::StopPump() ... mutex LOCKED

Could you please add an issue to the SDK repo with the above logs? This would help the SDK team to investigate and advise what could be wrong with this session. Thanks!!

azure.cognitiveservices.speech SpeechRecognizer canceled automatically

Activity