Human activity detection using sparse representation

Human activity detection from videos is very challenging, and has got numerous applications in sports evalution, video surveillance, elder/child care, etc. In this research, a model using sparse representation is presented for the human activity detection from the video data. This is done using a linear combination of atoms from a dictionary and a sparse coefficient matrix. The dictionary is created using a Spatio Temporal Interest Points (STIP) algorithm. The Spatio temporal features are extracted for the training video data as well as the testing video data. The K-Singular Value Decomposition (KSVD) algorithm is used for learning dictionaries for the training video dataset. Finally, human action is classified using a minimum threshold residual value of the corresponding action class in the testing video dataset. Experiments are conducted on the KTH dataset which contains a number of actions. The current approach performed well in classifying activities with a success rate of 90%.