Human Motion Prediction Based On Deep Learning

Computers' knowing how to predict human's motion is a important way for humancomputer interaction.It makes computers have an ability to act according to what a man is going to do,from which computers can do something in time.How to design the calculation,which makes computers' prediction conforming to human kinematics,is a difficult point.Deep learning is a characteristic learning method,the raw data can be changed to more abstract semantics by a deep neural network model.Motion prediction can be realized by predicting the semantics and reconstructing the motion.So the goal of this project is to design a neural network model which can predict human's motion and have a good construction result.Based one this goal,our work include analysing nowadays researching result about human motion predicton,doing some research about deep learnin,and method on dealing with human motion data.Our work also put forward two method to increase accuracy on the exsisting method.With some experiment,it can be proved that our method is useful.Finally,our work realizes the reconstruction of human motion in Unity3 D software.