Research on Human Motion Data Filtering Based on Unscented Kalman Filter Algorithm

With the continuous development of economy in China, manual production has long been unable to meet the current social needs, and more and more companies have begun to use robots instead of manual operations to carry out changes[1]. In the manufacturing field, human-machine collaborative manufacturing can effectively combine the advantages of industrial robots and workers, reduce production costs, improve product quality, make manufacturing more flexible and individualized, and develop in the direction of high speed, high precision and high efficiency. A key factor in the manufacturing of human-machine collaboration is the safety and security, namely industrial robots operate with a staff job at the same time in the working environment, not harm, directly or indirectly to the staff. Therefore, how to ensure that the industrial robot can accurately obtain the current position of the staff and predict the future position, ensure the safety of the staff, and clarify the intention of the staff is an urgent problem to be solved. In this article, a method aimed at improving the accuracy of Kinect v2 sensor skeleton data is proposed. This method uses the kinematic filtering method of unscented Kalman filter to try to achieve consistency by combining observation data with the motion prediction data being executed. Finally, the Kinect v2 SDK is corrected by extracting motion information and constraining the length of human bones. Some errors occurred in the measurement data. The data source of this experiment is measured by Kinect v2 SDK. The algorithm is realized by Python, and the result of this experiment is compared with the data of Kinect v2 SDK.