I went through the steps to test the tutorial but found that the performance of Julius was considerably inferior to what was posted here. I have not been able to get above %50 word error rate.
In particular, the recognizer seems to have lots of difficulty in identifying a long stream of digits. It is only able to recognize the first few digits. My guess is that the settings in the Julian.jconf file might need changing. Any suggestions?
Hi DRF,
Sorry for the delay in getting back to you on this, I am travelling...
> the recognizer seems to have lots of difficulty in identifying a long
>stream of digits
Yes, I have noticed the same thing - that is why the tests were limited to digit sequences of about 5-7 digits long.
You need a better acoustic model - i.e. use more audio in you're acoustic model.
Ken