The recognition of moving human body posture based on combined neural network

Abstract

A new method based on combined neural network is presented for the recognition of moving human body posture. The silhouette feature, skeleton feature and moment invariant feature of human body posture are firstly extracted, and every feature vector is inputted into their own neural network classifiers, then the outputs of all the classifiers are fused together with the Dempster-Shafer theory to form a combined neural network, so that a more powerful classifier with high recognition rate can be built. The experimental results show that the proposed method is more accurate than single neural network classifier for the recognition of moving human body posture.