Atefe Aghaei - University college of Rouzbahan, Sari ,Iran
Sajjad Tavassoli - Department of Computer Engineering, Sari branch, Islamic Azad University, Sari, Iran

Human action recognition is an important problem in computer vision. One ofthe methods that are recently used is sparse coding. Conventional sparse codingalgorithms learn dictionaries and codes in an unsupervised manner and neglectclass information that is available in the training set. But in this paper for solvingthis problem, we use a discriminative sparse code based on multi-manifolds. Wedivide labeled data samples into multi-manifolds and also to decrease run timereduce dimension of manifolds. We find k inter nearest neighbors and intra nearestneighbors for each data sample in each manifold. The intra class variance should beminimized while the inter class variance should be maximized, in the result we couldcalculate laplacian matrix and optimize sparse code and dictionary. Then we usediscriminative sparse error for classification. We run this method on KTH and UCFsport datasets. Results show that we obtain a better result (about 89%) in UCFdataset.

Action recognition, Discriminative Sparse Representation, Multi Manifold,Spatio-Temporal descriptors, Neighborhood