| Motion capture is one of the basic researches on wearable computing,humancomputer interaction,and ubiquitous computing.It is a cross-cutting field that combines multiple disciplines and technologies and has broad application prospects in sports training,medical health,film and television production,and other fields.Compared with optical,electromagnetic,mechanical,and other motion capture technologies,inertial motion capture has the advantages of not being affected by light,being restricted by the site,and easy to wear.It has become a hot spot for research and application by enterprises and scholars.However,in inertial motion capture,the sensor is susceptible to noise interference,the accuracy is not high,the acceleration estimation displacement has a cumulative error,and the commercial capture system has problems such as expensive equipment and no open source.Therefore,this thesis studies the inertial human body posture and displacement motion capture.The main work contents and innovations are as follows:1.This thesis builds a software and hardware platform for inertial human body posture and displacement motion capture.The hardware system mainly includes nineaxis inertial sensors(accelerometers,gyroscopes,and magnetometers),routers,and PCs.The wireless communication between multi-sensor nodes and PCs is realized through a communication protocol designed independently.The host computer software receives the raw data synchronously transmitted by the sensor in real-time through the data acquisition module,imports the character model after data processing,and finally realizes the capture of human body posture and displacement action.2.This thesis studies the principle and method of sensor calibration,analyzes the posture fusion algorithm of commonly used sensors,and proposes a new human body calibration method.According to the different characteristics of accelerometers,gyroscopes,and magnetometers,different attitude fusion algorithms are studied and analyzed,including complimentary filter fusion algorithms,gradient descent fusion algorithms,an Extended Kalman Filter(EKF)fusion algorithms.In the theoretical analysis and experimental comparison of different attitude fusion algorithms.The experimental data shows that the accuracy of the EKF algorithm and the gradient descent algorithm is slightly higher than that of the complementary filter algorithm,but in terms of time efficiency,the gradient descent algorithm is superior to the EKF fusion algorithm,so this paper uses a gradient descent-based data fusion algorithm.Finally,a calibration method that uses the world coordinate system as the reference system is proposed to achieve the coordinate matching problem between the human body and the host computer model.3.A single-node yaw correction model and a multi-node kinematic displacement model for lower limbs are proposed.Using the inertial measurement unit fixed on the instep of the foot,based on the zero-speed update algorithm(Zero-Speed Update,ZUPT)and EKF algorithm,it is proposed to correct the yaw angle error during human body movement to achieve the average error is a 5.35%.The multi-node kinematic displacement model is based on the principle of human kinematics,and a new squat calibration method is used to align the initial coordinate system of the human body,and then proposes a support leg detection algorithm to determine the contact between the human leg and the ground.Taking the contact point between the leg and the ground as the root node,the kinematics theory is used to accurately estimate the displacement of the center of mass of the human body.The experiment results are obtained by walking different routes.It shows that the average error is less than 4%.Finally,using the built software and hardware platform and the researched algorithm,the character model in the software of the host computer is driven by data to capture the posture and displacement changes of the human body,which further verifies the effectiveness and feasibility of the algorithm proposed in this paper. |
PDF Full Text Request