i have implement speech recognition using pocketsphinx in android application, works fine except recognizer sensitive. made application result (speech-to-text) not working properly.
i have change threshold until lowest can get, 1f.
i used english, , limited library grammar, , make own dictionary using sphinx knowledge base tool.
even though used acoustic model en-us-ptm demo of cmusphinx.
i have done research sphinx have remove_noise function, can seems find it. or there other way make pocketsphinx recognizer better ?
thankyou.
latest pocketsphinx has internal noise reduction module. there no easy way can remove/reduce noise have done if there such.
external noise reduction degrades speech recognition accuracy. make sure you're using updated version of library , language model
see more increasing accuracy of recognizer: http://cmusphinx.sourceforge.net/wiki/faq#qwhy_my_accuracy_is_poor
Comments
Post a Comment