Hola, acabo de descomprimir el quickstart de Julius y lanzo el run_julian.bat (bajo Windows) y todo perfecto, pero cuando cambio el .jconf y le pongo " -input mfcfile " y "-filelist audiodump.htk" para que lea el archivo.
La linea de comando que ejecuto es esta:
>> julian -C julian.jconf
Me sale un mensaje con "No such file or directory"en la funcion fopen_readfile.
Adjunto una traza de la linea de comandos en txt.
A ver si me podeis ayudar.Gracias.
Hello, I just unzip the quickstart of Julius and released the
run_julian.bat (under Windows) and everything perfect, but when the
change. Jconf and put "mfcfile-input" and "-filelist audiodump.htk" to
read the file .
The command line executed is:
>> Julian-C julian.jconf
I get a message "No such file or directory in fopen_readfile function.
Attached a trace of the command line in txt.
Let's see if I can ayudar.Gracias.
>but when the change. Jconf and put "mfcfile-input" and "-filelist
>audiodump.htk" to read the file .
Have you used the files listed in audiodump.htk successfully with HTK? Which config file did you use to create the feature sets listed in audiodump.htk? The Julius config assumes the following HTK config format:
SOURCEFORMAT = WAV
TARGETKIND = MFCC_0_D
TARGETRATE = 100000.0
SAVECOMPRESSED = T
SAVEWITHCRC = T
WINDOWSIZE = 250000.0
USEHAMMING = T
PREEMCOEF = 0.97
NUMCHANS = 26
CEPLIFTER = 22
NUMCEPS = 12
Hi Ken, thanks for your answer.
i don't know the feature sets for audiodump.htk
Audiodump.htk is as file generated with sox
>>sox mifile.wav audiodump.htk
is all information that i have.
How i can know the settings for a HTK file?
Best regards from Valencia(Spain)
P.D. My name is Raúl, and not "Ayudar", google is not the best translator ! ;.) "Ayudar" means, it's the same as verb "Help"
>Audiodump.htk is as file generated with sox [...]
You might try creating your HTK files using HTK's HCopy command (see Step 5 of the VoxForge tutorial).
>How i can know the settings for a HTK file?
The HTK HList command can be used to display the header of a file. See the HTKBook for details, section 5.15 Viewing Speech with HList.
But rather than trying to convert from WAV to HTK format, and trying to recognize in HTK format, why not work directly with WAV format file (using the "-input rawfile" parameter)?