Speech Recognition Engines

Nested
Julius with baby monitors
User: pomprocker
Date: 2/12/2010 6:22 pm
Views: 6210
Rating: 6

I'm trying to use a baby monitor receiver connected to the input on my sound card on a linux server in the garage and the transmitter in my room upstairs as a 'room mic'. The mic is so sensitive it even picks up the ticking of my wall clock. I have even set the receiver to the lowest volume for julius. I am having some problems as you can imagine, Julius thinks everything is input:

 

<<< please speak >>>WARNING: adin_thread_process: too long input (> 320000 samples), segmented now

So is there a way to adjust Julius' sensitivity to deal with this?

 

 

 

--- (Edited on 2/12/2010 6:22 pm [GMT-0600] by pomprocker) ---

--- (Edited on 2/12/2010 6:23 pm [GMT-0600] by pomprocker) ---

Re: Julius with baby monitors
User: kmaclean
Date: 2/17/2010 9:42 pm
Views: 2685
Rating: 7

>So is there a way to adjust Julius' sensitivity to deal with this?

From the Julius manual, you might try:

       -lv threslevel
              Level  threshold  (0  -  32767) for speech triggering.  If audio
              input amplitude goes over this threshold for  a  period,  Julius
              begin  the  1st  pass recognition.  If the level goes below this
              level after triggering, it is the end  of  the  speech  segment.
              (default: 2000)

 

--- (Edited on 2/17/2010 10:42 pm [GMT-0500] by kmaclean) ---

PreviousNext