基于少量惯性传感器的姿态重建与仿真

发布时间：2018-05-24 16:00

  本文选题：运动捕捉 + 姿态重建　； 参考：《系统仿真学报》2017年10期

【摘要】：提出了一种基于少量惯性传感器的姿态重建方法,将输入的低维角加速度信息作为控制信号,以高维欧拉角运动数据库为基础,重建出高维的人体全身姿态序列。通过建立数字相似性-几何相似性-时间持续性模型,对加速度进行重构,保证候选重建姿态的在数字-逻辑上相似性的统一。通过引入最优化能量函数,求解最优重建姿态序列。通过分析和对比实验,表明方法能够重建出真实可信的人体姿态序列。本研究使用少量惯性传感器控制,能产生持续的全身关节的姿态序列,通过减少传感器数量,达到降低系统成本的目的。
[Abstract]:An attitude reconstruction method based on a small number of inertial sensors is proposed. The input low-dimensional angular acceleration information is taken as the control signal, and the high-dimensional human body pose sequence is reconstructed based on the high-dimensional Euler angle motion database. By establishing the model of digital similarity, geometric similarity and time persistence, the acceleration is reconstructed to ensure the unity of digital-logic similarity of candidate reconstruction posture. The optimal reconstruction attitude sequence is solved by introducing the optimal energy function. Through analysis and contrast experiments, it is shown that the method can reconstruct the true and credible human pose sequence. In this study, a small number of inertial sensors are used to control the attitude sequence of the whole body joint, which can reduce the cost of the system by reducing the number of sensors.
【作者单位】： 北京理工大学软件学院;空军装备部107办公室;中国航空工业集团成都飞机设计研究所;航天系统仿真重点实验室;中国航天科工集团公司;
【基金】：国家自然科学基金(61202243)
【分类号】：TP212.9
，

本文编号：1929687