Enhanced human behavior recognition using HMM and evaluative rectification

Abstract

Human behavior recognition and real world environments monitoring constitute challenging research problems rapidly gaining momentum over the last years. Methods for time series classification like the Hidden Markov Models have been employed in the past for similar tasks, however in many challenging cases they fail, since some behaviors are much more difficult to model than others. This happens particularly in cases that there is scarcity of labelled data. In this paper we introduce a novel re-adjustment framework of behavior recognition and classification by allowing the user incorporation in the learning process. The proposed Evaluative Rectification approach aims at dynamically correcting erroneous classification results to enhance the behavior modeling and therefore the overall classification rates. We evaluate the performance of the examined approach in a challenging real-life industrial environment of an automobile manufacturer. Our experiments indicate a significant outperformance of the proposed Evaluative Rectification scheme compared with traditional classification frameworks, such as Hidden Markov Models.