How to use Matlab HMM toolbox to do classifcation

I want to use HMM to do some multi-label classification on my data set.

The train data has about 10000 samples. Each sample is a vector that has 100 elements. And the train label is a number from 1 to 10.

The test data has about 1000 samples, with the same format.

Now I want to do HMM which is a supervised learning method on my data set, and use them to predict on the test, to calculate the error rate.

I am using Matlab, so I want to find a toolbox to use. I am going to use Kevin Murphy's Matlab toolbox. But I am not sure which model in his toolbox should I use:

HMMs with discrete outputs or HMMs with mixture of Gaussians outputs?

How to get a HMM model? And use which algorithm to predict the label by using the test data?

I did not quite understand his tutorial on his website. Could anyone give me some sample code?

Thanks!