With the development of science and technology and the improvement of people's quality of life, film and television animation has gradually become an important means to enrich people's daily entertainment and cultural exchanges. The inertial motion capture technology developed on the basis of motion capture has accelerated the development process of film and television animation, making film and television animation effects more colorful and precise in place. This paper presents a method of using inertial sensors to capture human arm motion. Using the quaternion information calculated by the inertial measurement unit, the positions of the joint points of the arm wrist, elbow and shoulder are obtained. The inertial data are transmitted to the host computer through the Bluetooth wireless communication method. The OptiTrack optical motion capture device with millimeter-level motion capture accuracy is used to obtain the position data of the human arm and use it as the reference position data. This paper analyzes the source of the zero-speed detection misdetection combined with experiments, and optimizes the detection algorithm based on the Laida criterion. On this basis, a complete trajectory capture Kalman filtering algorithm framework is constructed. Its embedded attitude depth correction module can effectively suppress the attitude of the angular error diverges. Finally, through physical experiments, the feasibility of the trajectory capture algorithm is verified in terms of accuracy and stability. The test shows that the trajectory capture algorithm described in this article can obtain a higher smoothness of the walking trajectory, so it can reduce the animation complex. The shaking of the human body model during the present process has important engineering significance.

随着科技的发展和人们生活质量的提高，影视动画逐渐成为丰富人们日常娱乐和文化交流的重要手段。在动作捕捉的基础上发展起来的惯性动作捕捉技术，加速了影视动画的发展进程，使影视动画效果更加丰富多彩，到位精准。本文介绍了一种使用惯性传感器来捕捉人体手臂运动的方法。利用惯性测量单元计算出的四元数信息，得到手臂腕、肘、肩关节点的位置。惯性数据通过蓝牙无线通信方式传输到上位机。采用毫米级运动捕捉精度的OptiTrack光学运动捕捉设备，获取人体手臂的位置数据，并将其作为参考位置数据。本文结合实验分析了零速检测误检的来源，并基于莱达准则对检测算法进行了优化。在此基础上构建了完整的轨迹捕捉卡尔曼滤波算法框架。其内嵌的姿态深度校正模块可以有效抑制姿态角误差发散。最后通过物理实验验证了轨迹捕捉算法在准确性和稳定性方面的可行性。测试表明，本文所述的轨迹捕捉算法可以获得较高的行走轨迹平滑度，所以它可以减少动画复杂度。当前过程中人体模型的抖动具有重要的工程意义。