I have a problem in recognizing all english words using sphinx4. I added all the english words starting with letter ‘A’ to my grammar file in java and tried to run it. First time it gave me the error of java heap space which i initially kept 256m.
Then i changed it to 512m. this time there were no errors from java, but the recognizer was not recognizing a single word not even wrong words.
Can any one help me in making a speech recognition grammar file with all english words…
This is my first question here so hoping to get the some good results.
Since you are using SPhinx 4. A lot of the things depends on the way you have configured. Sphinx (both pocket sphinx and Sphinx 4) offers transcription as well as immediate conversion. I mean, you can save your voice as a wav file and perform a wav to text conversion or you can simply speak keeping the sphinx4 application running so it will transcribe every line that u speak.
You would have to configure the grammar file as in using a 2/3/n gram file. Provide the acoustic model based on the pronunciation as in hub4 acoustic model for US english or voxforge models for indian english. I had earlier posted a similar question
Speech recognition response is poor in sphinx4
You can also refer to a this site http://nsh.nexiwave.com/ which provides all the necessary details for configuring Sphinx4 . based on these configurations i suppose you should be able to get a dignified output