Action Recognition in Multimedia Streams

摘要: It is well accepted that the rise in proliferation of inexpensive digital media collection and manipulation devices has motivated need to access this data by content rather than keywords. The requirements based are understood research community there no elaborate further here. Parsing multimedia streams detection classification action implies modeling dynamic nature visual audio features as they evolve time. Hidden Markov Model (HMM) long been used model behaviour signals. Its power capture complex domain led widespread use analysis because non-stationarity inherenet those However, subtleties application HMMs often unclear framework processing latter portion chapter sets out expose some these. Three applications considered motivate discussions: actions sports, observational psychology illicit video content. Sports: Work sports understanding conducted for a decade now with clear motivation provided huge amount broadcasting on internet television. An overview footage general can be found [22]. Action recognition here involves certain plays situations dictated game e.g. pots, goals, wickets aces. Illicit Content: distribution pornographic materials also benefited from revolution [6]. This kind material illegal workplace referred article. issue filtering major concern since introduction web early 1990’s. Pixalert’s ‘Auditor’ ‘Monitor’, FutureSoft’s ‘DynaComm i:scan’ 2 Hyperdyne Software’s ‘Snitch’ all provide image