Traffic Light, and a crossraods of algorithms :)

My current dilemma with my PhD is choosing an engine for decision making. What decisions to make? Well... now I have a bunch of video segments, where each has specific soccer events in them. Now... since I've never used any machine learning or statistical methods in my work... I am facing difficulty even in starting to plan about the whole thing. What to use as features? How to train and test? Which engine to use? The Hidden Markov Models seems practical... but like I said, dunno where to start. The Support Vector Machine could also be a good candidate as if modeled properly, can solve non-linearly-separable problems quite easily. Then of course, there's the Bayesian Decision Theory... Bayesian classifiers seem simple to understand... but at one point, I get lost. Hahahaha! Anywayz... this is where I am now. I pray to God to help me search for what I should find... Ameen :D Okeh, meh sambung kerja...