This is a demo version of my code for the paper: “ Arousal Content Representation of Sports Videos
Using Dynamic Prediction Hidden Markov Models” just run “demo.m” . The code shows the predicted curve for every iteration of the EM algorithm and compares it to linear regression for a 2 state example. The function required are all included
I made some changes with respect to regularization so it runs smoothly for this data but it may not always converge to a local maximum so you may have to run it multiple times.
generates toy data as well.