Human action recognition based on the angle data of limbs
Maierdan Maimaitimin, Keigo Watanabe, Shoichi Maeyama · 2014
An approach to human action recognition is presented in this paper. This paper is part of an human behavior estimation system which is divided into two parts: human action recognition and object recognition. In this part, we use Microsoft Kinect to capture human joint data. And calculate the limb angles. Using these angles we can train an Artificial Neural Network(ANN) to recognize these actions, which in this case are "walking" and "running". In this paper, ANN is discussed as a main part of the current research. We designed a two stage ANN, which can minimize the impact of noise data. Whole processing is simulated by Scilab.