VoxForge
VoxForge
Hello there,
I have tried to deal with .wav format file in HTK under my windows and linux operation systems. Unfortunately, they are both failed......:"(
Could you please do me a favor to run a short program on your HTK? I really need the HTK can deal with .wav format files:"(
I have packaged all my experimental files and I can send them to you if you are so kind to do me a favor. It will just take you about 5 mins. You only need to run two commands on your HTK:
1. HCopy -T 1 -C config.conf -S codetr.scp
2. HCompV -C config.conf -f 0.01 -m -S train.scp -M hmm0 proto.pcf
I don't know how to upload my files here:"(
If you can help me , could you please tell me your email? I can send you all the files.
What you only need to do is to extract my files into your disk and run the two commands. Really, it will only take you 5 mins.
Thank you so much for your great help!!!
--- (Edited on 5/3/2012 4:52 pm [GMT-0500] by NancyZxll) ---
NancyZil, HTK does work with wav files as usuall. You can download any example from this website and try HTK with it and it will work fine.
As for sharing the files, you can use public sharing service like google docs, google drive or dropbox and just paste here a link to your data.
--- (Edited on 5/6/2012 12:05 [GMT+0400] by nsh) ---
Thank you so much nsh, I will download some sample audio files to try again.
My files are now all in https://www.dropbox.com/sh/n1zmk4pl5215xmb/JNuOcPGuWT
It's the dropbox downloading link, the HTK zip file.
Thank you so much!!
Just download it and run the above two commands are OK.
Thank you again!
--- (Edited on 5/6/2012 10:53 pm [GMT-0500] by NancyZxll) ---
Hello Nancy
Ok, I've got your trouble. The issue you have is that you share the same config between HCopy and HInit. They must be different config files
In HCopy config which you should name wav.config you should have a line
SOURCEKIND=WAV
in HInit config which you should name train.config you should have a different line
SOURCEKIND = MFCC_0
Because you are now trying to input the MFCC data, not wav data. All this is described in tutorial, just read it carefully.
Also, your file is 8khz. Make sure you intended this. Tutorial describes training with 16 khz.
If you want to save yourself from such a painful HTK training process I recommend you to try CMUSphinx:
http://cmusphinx.sourceforge.net/wiki/tutorial
It's way easier to train and use.
--- (Edited on 5/7/2012 11:59 [GMT+0400] by nsh) ---
You need to learn one lesson from this. The more information you provide and the more accurate is it the faster you resolve the problem. You could complain about WAV format for one more month.
--- (Edited on 5/8/2012 10:20 [GMT+0400] by nsh) ---