International Journal of Communication and Computer Technologies (IJCCTS)
In this paper human actions are recognized through Generalized Regression Neural Network (GRNN) and used for identifying particular actions. Self Organizing Maps (SOM) is used for the representation of action from the videos, which is based on learning spatially related human body posture prototypes. A time invariant action representation can be obtained using fuzzy distances from human body posture prototypes. For action classification Generalized Regression Neural Network (GRNN) is used. The algorithm is trained using data from multiple cameras.