Movement Pattern Histogram for Action Recognition and Retrieval

Abstract

We present a novel action representation based on encoding the global temporal movement of an action. We represent an action as a set of movement pattern histograms that encode the global temporal dynamics of an action. Our key observation is that temporal dynamics of an action are robust to variations in appearance and viewpoint changes, making it useful for action recognition and retrieval. We pose the problem of computing similarity between action representations as a maximum matching problem in a bipartite graph. We demonstrate the effectiveness of our method for cross-view action recognition on the IXMAS dataset. We also show how our representation complements existing bagof-features representations on the UCF50 dataset. Finally we show the power of our representation for action retrieval on a new real-world dataset containing repetitive motor movements emitted by children with autism in an unconstrained classroom setting.

Effective and efficient human action recognition using dynamic frame skipping and trajectory rejection

“…Draft for the preparation of submission to Image and Vision Computing of state-of-the art approaches [11,18,28,29,[33][34][35][36][37][38][39][40][41]. For the sake of simplicity, we only report the results for the case in which we combined all descriptors.…”

Improved Motion Description for Action Classification

“…Their approach learned the codebook in a supervised manner to further improve performance. Ciptadi et al (2014) introduced a novel way to represent an action as a set of movement pattern histograms that encode the global temporal dynamics of the action, slightly deviating from the above pipeline. Hoai and Zisserman (2014) proposed a technique that addresses temporal interval ambiguity of actions by learning a classifier score distribution over video subsequences.…”

Efficient and effective human action recognition in video through motion boundary description with a compact set of trajectories

“…In order to verify the effectiveness of the proposed approach, we compared our approach to IDTF [10] and other state-of-the-art approaches [14,17,22,[26][27][28][29][30][31]. In particular, for IDTF, we extracted histogram-based descriptors from volumes with a size of N × N × L in the video frames, where N and L denote the spatial size of a volume and the trajectory length, respectively, and where we set these parameters to a value of 32 and 15, respectively [10].…”

“…Our evaluation consists of three experiments: 1) an experiment that investigates the effectiveness of the proposed approach; 2) an experiment that investigates the efficiency of the proposed approach; and 3) an experiment that compares the proposed approach with state-of-the-art HAR approaches [10,14,17,22,[26][27][28][29][30][31]. We present our results in the following subsections.…”