VoxForge
VoxForge
Hey,
I'm wondering if anyone has any experience using Julius + HTK with Gaussian Markov Models (GMM). I'm looking for an example/tutorial anything really that explains how to approach this.
Many Thanks,
A
--- (Edited on 5/5/2010 10:59 am [GMT-0500] by abnev) ---
You mean Gaussian Mixture Model (GMM)!
Why do you want to use GMMs? So you don't want to do ASR with that?! Do u want to do event detection or so? Otherwise using just GMMs makes no sense.
You could try to use a HMM prototype which looks like this:
~o <VecSize> 39 <MFCC_0_D_A>
~h "proto"
<BeginHMM>
<NumStates> 3
<State> 2
<Mean> 39
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
<Variance> 39
1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
<State> 3
<TransP> 3
0.0 1.0 0.0
0.0 0.0 1.0
0.0 0.0 0.0
<EndHMM>
This should be similar to a GMM!
Greez
--- (Edited on 5/25/2010 4:52 am [GMT-0500] by Visitor) ---
You mean Gaussian Mixture Model (GMM)!
Why do you want to use GMMs? So you don't want to do ASR with that?! Do u want to do event detection or so? Otherwise using just GMMs makes no sense.
You could try to use a HMM prototype which looks like this:
~o <VecSize> 39 <MFCC_0_D_A>
~h "proto"
<BeginHMM>
<NumStates> 3
<State> 2
<Mean> 39
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
<Variance> 39
1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
<State> 3
<TransP> 3
0.0 1.0 0.0
0.0 0.0 1.0
0.0 0.0 0.0
<EndHMM>
This should be similar to a GMM!
Greez
--- (Edited on 5/25/2010 4:54 am [GMT-0500] by Visitor) ---