One place to start is an algorithm like Baum-Welch that uses expectation maximization to find most likely parameters. MATLAB implements a version of this algorithm in 'hmmtrain'.

you can use the artificial neural network toolbox in Matlab especially for pattern recognition , where you can denote to left and right lane by binary targets (1 and -1)

I'm very grateful for your answers.

@ Hencey : for using hmmtrain I need the initial matrix A (transition matrix) and B (emission matrix matrix) like : [estTR,estE] = hmmtrain(seqs,trans,emis) could you please tell me how I cann compute initial transition & emission matrix?

@ Ihsan A. Baqer: could you please tell me some details about ANNs? how can I use this command?

thanks alot in advanced for your help

I send to you the attachment file from my Ph.D thesis (Appendix A), I hope that help you and i ready to help you in your problem.

Hello Mr. Dr. Baqer,

Thank you very much for your answer and your help.

Could you please tell me how I should choose the Input & Target ? should I have labeled Data?

(I attached one measrement data, the parameter are 1. column: Time, 2. speed, 3. steering wheel abgle, 4- steering wheel angle velocity, 5-column, steering force.)

an input & Target should I use these Data? and together or individually?

thanks alot in advanced for your help

Ahad

Hello Mr. Ahad

can you send your data file to my mail ([email protected]) because it is failed to download from this site.

ihsan

Also look at a Kalman filter approach.   

Hello Mr. Ahad Khezerloo

were you able to get a hold of an answer in regards to your question about the initial transition and emission matrix for the hmmtrain?

I'm wondering the same thing and hoping I could find the answer here.

You can train HMM with Viterbi alghoritm too (also in Matlab).

And initial matrix values generate randomly.

Do you have any data It has been my question for 2 month

I need help please

How can I use kalaman filter to get the matrix constants and predict the path of the vehicle ?

THanks