search

VikParuchuri / scribe

Simple speech recognition using your microphone.
123 stars 47 forks source link

Accuracy issue #6

Open akshat9425 opened 6 years ago

akshat9425 commented 6 years ago

Hi,

i used your repo and CMU sphinx provided hindi language model,In very few audio files there is very less accuracy will you please tell me that how could i improve accuracy or what kind of speech file does it accepts for better accuracy

Thanks

akshat9425 commented 6 years ago

few files throws error: "result = decoder.hyp().hypstr" please let me know how do i resolve such kind of issues and hopefully it doesn't found desired things in file i think

akshat9425 commented 6 years ago

Discard all above i found all solutions to improve accuracy by the way thanks

akshat9425 commented 6 years ago

i created my own dictionary so i will be able to pass my own generated .lm and .dic file but don't know what to pass as hmm folder

and In terms of accuracy some words are not said but printed i followed below link

https://jackscodingcraic.wordpress.com/2014/09/01/pocketsphinx-improving-accuracy-with-language-models-and-dictionaries-raspberry-pi/