Thread Subject: need help on creation lambda model for HMM

Dear all,

I’m trying to create a lambda model for HMM training. I need to find A(state transition probabilities) and B( state observation probabilities).
I have a dataset which is a 100x42 matrix. Every 41 columns of this dataset is feature vectors and 42. Column is my target values.
To create the HMM model, I’ve decided to use statistical toolbox and it’s hmmestimate command.

[TRANS, EMIS]= hmmestimate(seq, states)

so TRANS is by A and EMIS is supposed to be my B. I thought seq as my dataset and states as the targets of my dataset which is the 42th column of every line.

So i tried this code:
states=data(1:100, 42);
[trans, emis]= hmmestimate(data, states);
and i get the error: input sequence and states must be the same length

Is there any advice on obtaining initial parameters of A, B and pi to create an HMM ?

Dear all,
I am doing my project in image processing(Image edge detection using Hidden markov chain model based on undecimated wavelets).Markov chain is formed by considering the pixels from every sub bands.Now my sequence size is [65536X3]
Help me how to obtain the transition and emission parameters.Tell me the steps to train the model.Plz help me......