Comparison of feed forward and cascade forward neural networks for human action recognition

Humans can perform an enormous number of actions like running, walking, pushing, and punching, them in multiple ways. Hence recognizing a human action from video is challenging task. In supervised learning environment, are first represented using robust features then classifier trained for classification. The selection does affect the performance recognition. This work focuses on comparison two structures neural network, namely, feed forward network cascade Histogram oriented gradients (HOG) histogram optical flow (HOF) used as representing actions. HOG represents spatial while HOF gives motion video. architectures compared based recognition accuracy. Well-known publically available datasets interaction detection testing. It seen that, applications, better results terms higher accuracy than Cascade network.