Artificial Neural Network-Based Prediction of Human Posture

The use of an artificial neural network (ANN) in many practical complicated problems encourages its implementation in the digital human modeling (DHM) world. DHM problems are complicated and need powerful tools like ANN to provide acceptable solutions. Human posture prediction is a DHM field that has been studied thoroughly in recent years. This work focuses on using a general regression neural network (GRNN) for human posture prediction. This type of ANN has advantages over others when incorporated in DHM problems like posture prediction. A new heuristic approach is also presented in this study to determine the GRNN parameters that lead to the best performance and prediction capability. The results are promising: a high success rate is obtained for predicting 41 outputs, which represent the upperbody degrees of freedom of a human model. This work initiates future focus on embedding GRNN to generalize human posture prediction in a task-based