question

DavaaGmail-7922 avatar image
0 Votes"
DavaaGmail-7922 asked romungi-MSFT answered

having problem with InitialSilenceTimeout reason from Cognitive Service, Speech to Text

We have a problem that many of our voice files cannot get recognized when we used SpeechRecognition. They mostly fail with InitialSilenceTimeout reason.
When we listen the voices, they don't sound any silence in the beginning and also there are couple other voice files get recognized correctly. So we don't think it is code issue but we cannot figure out why it fails for some voices. There is Conversation_Initial_Silence_Timeout property and we set it to some seconds, but that doesnt help.
Also if I test exact same voice files in SpeechStudio with same Resourcegroup, they pass the test and recognized.
Does anyone had any similar issue? would there be any property that we missed to configure in the code?

azure-speech
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

1 Answer

romungi-MSFT avatar image
0 Votes"
romungi-MSFT answered

@DavaaGmail-7922 Have you tried to set the following property too with the speech config?

 speechConfig.SetProperty(PropertyId.SpeechServiceConnection_InitialSilenceTimeoutMs, "45000");

This thread from the speech SDK repo details the different scenarios of timeout for silence. Segmentation timeout and max segmentation timeout are couple of other timeouts that are not exposed with the SDK and if they are reached with RecognizeOnceAsync then I think such behavior could occur. It also depends on the quality of audio but since the speech studio is able to recognize I think this could be an issue with SDK or setting the above property could help.




5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.