@inproceedings {pub2529,
	title = {sNN-LDS: Spatio-temporal Non-negative Sparse Coding for Human Action Recognition},
	author = {Thomas Guthier AND Adrian Sosic AND Volker Willert AND Julian Eggert},
	year = {2013},
	month = {June},
	abstract = {Current state-of-the-art approaches for human action recognition focus on complex local spatio-temporal descriptors, while the spatio-temporal relations between the descriptors are discarded. This bag-of-words (BOW) based approaches come with the cost of limited descriptive power, because class-specific mid- and large-scale topological information, such as body poses, cannot be represented. To overcome this restriction, we propose sparse non-negative linear dynamical systems (sNN-LDS) as a dynamic, parts-based spatio-temporal representation of local descriptors. We provide learning rules based on sparse non-negative matrix factorization to simultaneously learn both, the parts as well as the transitions between them. The algorithm is capable of learning class-specific movements on two one-shot-learning scenarios and on the challenging UCF-sports dataset our sNN-LDS combined with simple local features is competative with the state-of-the-art BOW-SVM approach. },
	publisher = {Springer},
	booktitle = {Artificial Neural Networks andProceedings of the 24th International Conference on Artificial Neural Networks (ICANN)},
	pages = {185-192}
}