Human Motion Posture Recognition In Basketball Based On Inertial Sensors

In the field of basketball,coaches develop training programs and evaluate the quality of training mainly through personal experience,which is limited.And it is subjective to use the way of artificial observation and evaluation.Body Area Network can assist athletes training and identify their sports posture,which is helpful for athlete and coaches.And basketball technical movement contains much complex basic action.And the accurate recognition of basketball posture is useful for basketball game and training practice.We apply the technology of inertial sensors to the posture recognition in basketball in this article which includes three aspects.First,we apply the Kalman filter to attitude algorithm and improve the accuracy of attitude algorithm to improve the accuracy of sensors and reduce the noise interference.Second,the human body model is constructed on PC to solve the limitation and complicated of optical motion capture in sports training.The lower limb movement is initially realized in this model.Third,based on the above research,this paper propose the recognition method of basketball posture,which is according to the motion information from sensors on legs and arms.The action status of legs and arms is detected by sensors,and the signal is segmented accurately to reduce the influence of other extraneous signal and improve the accuracy of recognition.This paper analyzes the behavior of the legs and arms movements and the characteristics of signal waveform of the nine kinds of movements in walking,running,jumping and standing dribbling,walking dribbling,running dribbling,shooting,passing and catching.The two-stage data segmentation method of motion is proposed,and the eigenvector is extracted.We construct classifiers for each motion with the four machines learnings to achieve the recognition.Experiments show that the method of basketball posture recognition can recognize nine kinds of sports postures in basketball,and the average accuracy rate of different testers can reach 98.85% which has certain practical value.