julius-speech / julius

Open-Source Large Vocabulary Continuous Speech Recognition Engine
BSD 3-Clause "New" or "Revised" License
1.84k stars 300 forks source link

breathing sounds when using a grammar model for speech recognition #194

Open alexkinren opened 1 year ago

alexkinren commented 1 year ago

"I am using Julius for speech recognition with the following grammar configuration: julius -C /usr/local/src/julius/julius-kit/dictation-kit/am-gmm-monophone.jconf -walign -nostrip -1pass -gram /usr/local/src/julius/dict_test/test -input file. I've encountered an issue. When there is a breathing sound lasting about 500ms in the audio file, all the subsequent speech is recognized as [/s]. Do you have any solutions for this?