Automatic human action recognition in videos by graph embedding

Publication Type:
Conference Proceeding
Citation:
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2011, 6979 LNCS (PART 2), pp. 19 - 28
Issue Date:
2011-09-26
Full metadata record
The problem of human action recognition has received increasing attention in recent years for its importance in many applications. Yet, the main limitation of current approaches is that they do not capture well the spatial relationships in the subject performing the action. This paper presents an initial study which uses graphs to represent the actor's shape and graph embedding to then convert the graph into a suitable feature vector. In this way, we can benefit from the wide range of statistical classifiers while retaining the strong representational power of graphs. The paper shows that, although the proposed method does not yet achieve accuracy comparable to that of the best existing approaches, the embedded graphs are capable of describing the deformable human shape and its evolution along the time. This confirms the interesting rationale of the approach and its potential for future performance. © 2011 Springer-Verlag.
Please use this identifier to cite or link to this item: