VoxForge
VoxForge
hello,guys, i just managed to build a simplest yes&no recognition system using HTK, but i still don't know how to train the HTK with a large corpus, like TIMIT, is there any tutorial online or who can give me something talking training HTK with a large corpus
--- (Edited on 8/22/2012 2:16 am [GMT-0500] by xw920) ---
Six years ago I wrote a single script that trained up phone recognition on TIMIT (TIMIT is not suitable for word recognition). If you are interested you can still get it at http://www.cantabResearch.com/HTKtimit.sh
But VoxForge is a large corpus, why not use that?
http://www.voxforge.org/home/dev/acousticmodels
Tony
--
Dr Tony Robinson
Founder Cantab Research Ltd
--- (Edited on 22-August-2012 8:27 am [GMT+0100] by TonyR) ---
Tony,thank you for replying, I'm a beginner of HTK, cuz i need to use HTK in my thesis, maybe the corpus i'll use is the IEEE corpus, so that's why i want to find some tutorial talking about how to train HTK with a corpus and something like that. because i think i'm still not familiar with HTK, for the yes&no recognition system i build, it is very simple, i just record my own voice(yes,no) , but with a corpus , i don't know how to start
--- (Edited on 8/22/2012 2:43 am [GMT-0500] by xw920) ---
Well the first place to start is the HTK book. Read through that then step through my scripts and Ken's scripts in conjunction with the manual and you should then have a good idea as to how everything fits together.
Tony
--
Dr Tony Robinson
Founder Cantab Research Ltd
--- (Edited on 22-August-2012 8:51 am [GMT+0100] by TonyR) ---
