Predicting viewer-perceived activity/dominance in soccer games with stick-breaking HMM using data from a fixed set of cameras

摘要: We attempted to predict activity/dominance for soccer games, where activity is defined as the degree of game perceived by viewer, whereas dominance at which viewer perceives a particular team dominate over other team. Such information would help layman understand game. It also enable construction an automatic digest creation system that extracts scenes having high activity/dominance. There are two facets this study: 1. The main part underlying prediction model consists Stick-Breaking Hidden Markov Model, data automatically estimates number states process behind data. 2. used in paper vector time-series consisting player, referee, and ball positions, together with information, acquired set fixed cameras. problem was approached Bayesian framework learning were implemented three different methods: Chain Monte Carlo, Expectation Maximization, Variational Bayes. proposed method tested using dataset 10 professional games compared against standard regression methods.