Research On Multi-node Motion Capture Based On Inertial Sensor

Human motion capture has a promising prospect in the field of athletics,animation and healthcare.Inertial motion capture is a technique that records human motion through inertial sensor.This technique has no site-limitation and is convenient for installation.The existing inertial motion capture products are expensive and the realization algorithms of this technique are strictly confidential,which dramatically influence the application and study of this technique.This dissertation makes a research on the technology of multi-node motion capture and the main contributions are as follows:1.The composition of a typical motion capture system is introduced.The theoretical foundations of calculating the rigid space gesture are analyzed.After comparing the differences between the Mahony algorithm and the Madgwick algorithm,this dissertation chooses the Madgwick algorithm with better performance.2.To display the outputs of the motion capture system,human body is abstracted as a model which contains 10 nodes in this dissertation.These 10 nodes contain the head,trunk and limbs,which can capture the common movements of the human body except gestures.Due to the inability to guarantee the consistency of each wearing position,all products of the inertial motion capture require a boot calibration.To simplify the multistep calibration procedures for commercial products,the single-step calibration algorithm is proposed.To visualize the capture results,a forward skeletal algorithm is put forward.Under the condition of ensuring the relative position of the node,this algorithm avoids the absolute coordinate of the node thereby reducing the computational load.3.An inertial motion capture experimental platform containing 10 nodes is built.This platform uses the MPU9250 as the primary sensor and summarizes the data through CAN bus thereby transmitting it to the PC via the wireless.Then,the multi-sensor fusion algorithm and the multi-node human motion capture algorithm are implemented on the PC side of the experimental platform.4.The power consumption and communication of the system are tested and the results show that the system can transmit all sensor data to the PC side stably.Then,this dissertation analyzes the characteristics of the raw data of the three sensors,thereby verifying the necessity of fusing three different sensors.After analyzing the values of three Euler angles when the magnetometer is started or not,this dissertation finds that the magnetometer has the functions of eliminating the heading drift and reducing the noises of roll and pitch.Eventually,after comparing the human action and the software animation,this dissertation concludes that this system can capture common human movements.