Action Classification with Locality-constrained Linear Coding
This work addresses action recognition for video analysis, presenting an incremental improvement over existing methods.
The paper tackled action classification in videos by using Locality-constrained Linear Coding (LLC) to capture discriminative information from spatiotemporal subsequences, achieving better average accuracy than ten state-of-the-art algorithms on five benchmark datasets.
We propose an action classification algorithm which uses Locality-constrained Linear Coding (LLC) to capture discriminative information of human body variations in each spatiotemporal subsequence of a video sequence. Our proposed method divides the input video into equally spaced overlapping spatiotemporal subsequences, each of which is decomposed into blocks and then cells. We use the Histogram of Oriented Gradient (HOG3D) feature to encode the information in each cell. We justify the use of LLC for encoding the block descriptor by demonstrating its superiority over Sparse Coding (SC). Our sequence descriptor is obtained via a logistic regression classifier with L2 regularization. We evaluate and compare our algorithm with ten state-of-the-art algorithms on five benchmark datasets. Experimental results show that, on average, our algorithm gives better accuracy than these ten algorithms.