The threshold between what the task identifies as silence and non-silence.
From the 11.2 release, HPE IDOL Speech Server Reference uses a new algorithm for audio preprocessing. The new algorithm uses DNN technology, which provides better performance and requires less tailoring to specific audio types.
The new implementation ignores speech and silence threshold parameters, and also distinguishes between music and noise, rather than simply recognizing music and noise as a single non-speech category. All tasks use the new algorithm, but the old algorithm is retained for backwards compatibility, and can be used in exactly the same way as before.
Action: | AddTask
CheckResources |
Task: | ClusterSpeech
ClusterSpeechTel ClusterSpeechToTextTel LangIdFeature LangIdSegWav LangIdCumWav LangIdBndWav LangIdSegStream LangIdCumStream LangIdBndStream SpeechSilClassification StreamToTextMusicFilter StreamToTextMusicFilterPunct TelWavToText TelWavToTextPunct |
Type: | Integer |
Default: | 50 |
Range: | 20–100 |
Example: | SilThresh=75
|
See Also: | MaxSilThresh (configuration parameter) |
|