Speech Recognition Engines

Flat
Can HTK only process the audio files generated by HSLab (the .sig format)?
User: NancyZxll
Date: 3/26/2012 1:08 pm
Views: 14263
Rating: 5

Hello there,

   I have recorded my voice and generated "wav" files by using Matlab "wavrecord" function and Windows "Sound Recorder".

   However, the HCopy cannot take useful wav file information out and then lead to HCompV generating errors.

   I wonder whether HTK can only process the audio files generated by HSLab in the sig format?

   Since I need to do a comparison experiment, I need wav audio files. 

   I am so appreciated your your help!!!

--- (Edited on 3/26/2012 1:08 pm [GMT-0500] by NancyZxll) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: kmaclean
Date: 3/26/2012 2:10 pm
Views: 58
Rating: 7

> I wonder whether HTK can only process the audio files

>generated by HSLab in the sig format?

Audacity works to create wav files that can be processed by the HTK tools

--- (Edited on 3/26/2012 3:10 pm [GMT-0400] by kmaclean) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: NancyZxll
Date: 3/26/2012 3:05 pm
Views: 117
Rating: 7

Hello Kmaclean,

    Thank you for your reply!! It's very usefull.

    Does Audacity generate a different format wav file from other softwares? I need to further transfer the wav files into spectrograms. I wonder whether the differences will lead to different spectrograms.

    Thanks a lot!!

--- (Edited on 3/26/2012 3:05 pm [GMT-0500] by NancyZxll) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: kmaclean
Date: 3/26/2012 9:32 pm
Views: 110
Rating: 5

>Does Audacity generate a different format wav file from other softwares? 

Don't really know... all I know is that it works with HTK tools...

--- (Edited on 3/26/2012 10:32 pm [GMT-0400] by kmaclean) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: NancyZxll
Date: 3/28/2012 10:27 am
Views: 78
Rating: 5

Ok, Thank you so much Kmaclean~~

--- (Edited on 3/28/2012 10:27 am [GMT-0500] by NancyZxll) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: NancyZxll
Date: 3/28/2012 1:54 pm
Views: 90
Rating: 8

Hello Kmaclean,

    I have recorded my voice using Audacity, and export it as WAV format.

    Unfortunately, the HCopy cannot recorgnize the wav file as a waveform.

    I debugged the HCopy, and found that when running "IsWave" function, it jumpped to the end without doing 'if' sentences. That is, the wav file was not recognized as a wav file.

   Did I set wrongly?

   I am so aprreciated your great help!!

   Many thanks!!

--- (Edited on 3/28/2012 1:54 pm [GMT-0500] by NancyZxll) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: kmaclean
Date: 3/28/2012 4:49 pm
Views: 60
Rating: 7

>    HCopy cannot recorgnize the wav file as a waveform.

are you setting your config file correctly? 

i.e. with a  SOURCEFORMAT = WAV

--- (Edited on 3/28/2012 5:49 pm [GMT-0400] by kmaclean) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: NancyZxll
Date: 3/28/2012 5:32 pm
Views: 74
Rating: 7

Hello Kmaclean,

    My configure file is like this:

--------------------------------------

SOURCEKIND = WAVEFORM

SOURCEFORMAT = WAV

TARGETKIND = MFCC_0

TARGETRATE = 100000.0

SAVECOMPRESSED = T

SAVEWITHCRC = T

WINDOWSIZE = 250000.0

USEHAMMING = T

PREEMCOEF = 0.97

NUMCHANS = 26

CEPLIFTER = 22

NUMCEPS = 12

ENORMALISE = F

------------------------------------------

And the wav file is 8kHz, 16-bits, mono, PCM format.

The device toolbar settings in Audacity are:

Windows DirectSound;Realtek HD Audio output; Realtek HD Audio Input:Mic Volume; 1(Mono) Input Channel.

 

When I run HCopy in command line, no error happened. However, when I run HCompV, it said 

-----------------------------------------------------------------

J:\QA\NameRecognition\samples\ZxllHTK>HCompV -C config1.conf -f 0.01 -m -S TrainAudacity.scp -M hmm0 proto.pcf

  ERROR [+6251]  Input file is not in RIFF format

  ERROR [+6213]  OpenWaveInput: Get[format]HeaderInfo failed

  ERROR [+6313]  OpenAsChannel: OpenWaveInput failed

  ERROR [+6316]  OpenBuffer: OpenAsChannel failed

  ERROR [+2050]  LoadFile: Config parameters invalid

------------------------------------------------------------------

 

So I debugged the HCopy command, and found that it cannot recognize my wav file. That is, the value of IsWave is FALSE.

 

I have met this problem third times, the same errors happened before when I use the wav files generated by Matlab and Windows 'Sound Recorder'. I doubt the errors existed because of wav files, not record equipments.

 

If I can only recorded my voice as "sig" format for HTK processing, I have double recording work to do for considering my comparison experiment: one is recorded by HSLab (HTK processing), the other is by Audacity (generating wav file). Considering my training data is large, the double recording work is too much... So I hope that wav file can work...

Could you please help me ?  Or ,could you give me some suggestions? Thanks a lot!!

--- (Edited on 3/28/2012 5:32 pm [GMT-0500] by Visitor) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: kmaclean
Date: 3/28/2012 7:08 pm
Views: 90
Rating: 7

>So I debugged the HCopy command, and found that it

>cannot recognize my wav file. 

Try following the VoxForge tutorial... get that working.  Then add some of your audio to the Voxforge tutorial samples, and get that working.  Then modify one thing at a time until you get the working training sequence you want.

HTK is very finicky if a step is not followed exactly, and does not give much info as to why it is not working - the problem can be at a much earlier step.

 

--- (Edited on 3/28/2012 8:08 pm [GMT-0400] by kmaclean) ---

Re: Can HTK only process the audio files generated by HSLab (the .sig format)?
User: NancyZxll
Date: 3/29/2012 2:02 pm
Views: 75
Rating: 8

Ok, Thanks Kmaclean, I will try it.

--- (Edited on 3/29/2012 2:02 pm [GMT-0500] by Visitor) ---

PreviousNext