This is a generalized approach. The specifics would depend on the actual content of the video and the goals of your analysis.