ANALYSIS OF HUMAN ACTIVITY PREDICTION USING TEMPORAL SEQUENCE PATTERN

Abstract

Activity analysis and acceptance plays an important role in a advanced ambit of applications from assisted active to aegis and surveillance. Most of the accepted approaches for activity analysis and acceptance wait on a set of predefined activities and bold a changeless archetypal of the activities overtime. We represent an activity as an basic histogram of spatio-temporal features, calmly clay how affection distributions change over time. The above contributions of our plan include: A accepted framework is proposed to systematically abode the botheration of circuitous activity anticipation by mining banausic arrangement patterns; Probabilistic suffix timberline (PST) is alien to archetypal causal relationships amid basic actions, area both ample and baby adjustment Markov dependencies amid activity units are captured; The context-cue, abnormally alternate altar information, is modelled through consecutive arrangement mining (SPM), area a alternation of activity and article co-occurrence are encoded as a circuitous allegorical sequence; We as well present a predictive accumulative activity (PAF) to characterize the adequation of anniversary affectionate of activity. The capability of our access is evaluated on two beginning scenarios with two abstracts sets for each: action-only anticipation and context- aware prediction. Our adjustment achieves. above achievement for admiration all-around activity classes and bounded activity units. We adduce and present able methods for mining predictive patterns for both a banausic and banausic (time series) data. Our aboriginal adjustment relies on common arrangement mining to analyze the seek space. It applies a atypical appraisal address for extracting a baby set of common patterns that are awful predictive and accept low redundancy.