Hi Nick,

The Julian build of Julius was merged into Julius with release 4.0.  I think you only need to set the Grammar settings ("-v" and "-dfa") within the Julius config file to do the same thing as Julian used to do:

####
#### Grammar
####
#### "-gram", "-gramlist" can be used multiple times
#-gram gramprefix        # (comma-separated list of) grammar file prefix
#-gramlist txtfile        # text file containing grammar prefixes
#-dfa dfafile -v dictfile    # specify DFA and dictionary separately
#-nogram            # reset all grammar list already specified

You would also set the "-h" and "-hlist" settings as you normally would.

Ken

Thanks for your help!

I just ran the new julius version like I would do it with julian and it works!!!! :-)

Hm... now I have a curious new problem!

I trained a new speech recognition model and I tried to start it with julius-4.1.2 but I get these warnings on my screen which will not end:

...

------
### read waveform input
STAT: AD-in thread created
<<< please speak >>>Warning: strip: sample 0-1957 is invalid, stripped
Warning: strip: sample 0-2048 is invalid, stripped
Warning: strip: sample 0-2046 is invalid, stripped
Warning: strip: sample 0-2047 is invalid, stripped
Warning: strip: sample 0-2047 is invalid, stripped
Warning: strip: sample 0-2047 is invalid, stripped

...

The curious thing about that is that my speech recognition model works with the julian command from julius-3.5.2 without any warnings!

Hope anyone has an idea where these warnings come from!

Regards, Nick

Ok, I found something to this topic for all those who have the same problem:

"The warning "Warning: strip: sample xxx-xxx is invalid, stripped" means that Julius has found a certain length of zero sample sequence at the audio stream and skipped them, since feature extraction may be confused by such input. The stripped part is always a silent part and does not affect the recognition result, but you can also disable this function by specifying "-nostrip" option to Julius."

So, I tried the additional "-nostrip" option which really eleminates the warnings but after <<< please speak >>> julius doesn't recognise any speech input anymore! So a new problem occures in my case...

... haha, I just found something that can solve the problem! It seems like if the warning massage has something to do with the number of chanel inputs - anyway, when I use the OSS interface (i.e. use "-input oss" instead of "-input mic") then everything works fantasic!

Regards!