Hidden Markov Model Training for Dynamic Gestures?

Question

I know there is a lot of material related to hidden markov model and I have also read all the questions and answers related to this topic. I understand how it works and how

Answer 1

What does the 0 mean in your model ? It seems to me in your data there are no direct transitions for both states, it always goes back to the state 0. Try something like the following in your data for a state transition sequence.

• 1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4 4
• 1 2 3 4 0 0 0 0 0 0 0 0 0 0 0 0 0
• 0 0 0 0 1 1 2 2 3 3 4 4 0 0 0 0 0
• 4 4 3 3 2 2 1 1 0 0 0 0 0 0 0 0 0
• 4 4 4 3 3 3 2 2 2 2 2 1 1 1 1 1 1

As a general rule:

I would recommend to work with openCV only after you have a proof of concept in Matlab/octave. This has two reasons. First of all you know exactly what you want to do and how it works, and don't waste your time implementing and debugging your theory in a 'low' level language (compared to matlab). Debugging algorithms in openCV is really time-consuming.

Secondly after you know your stuff works as expected, if you implement it and hit a bug (of openCV or C++, python) you know it's not your theory, not your implementation, it's the framework. It happened to me already two times that employed computer scientists implemented directly from a paper (after being told not to do so), spending 80% of the remaining time to debug the algorithm without ANY success only to find out that: they didn't really get the theory or some submodule of openCV had a slight bug which degenerated their results.

The link you've mentioned uses a HMM toolbox in matlab. Try to implement and understand your problem there, it's really worth spending the time. Not only you can verify each step for correctness, you can use the itermediate matrices with your openCV code after you have a working model.