基于捷联惯导的运动轨迹跟踪技术研究
发布时间：2018-07-30 07:51
【摘要】：越来越多的应用都需要对工作在室内和室外的人进行跟踪定位,在医疗,游戏、虚拟现实等领域,获取人体运动过程中的数据并进行跟踪和重建,可以对人体运动进行描述以获得更多潜在的信息。在临床康复治疗中,跟踪室外环境下病人的运动状况,可以使得病人的活动被实时监控并得到及时改正;在游戏领域,创造出逼真的虚拟世界,使其中的人物具有逼真的运动效果是动画制作的成功关键;在人际交互领域,智能机器不仅需要通过语音与人直接交流,还需要有运动,姿态等更加自然的交互方式;在更多现实领域内,人们期望获得日常生活中运动的时间长短,步行、奔跑的轨迹和运动的规律来进一步对自身活动规律进行调整。因此,实时记录,无线传输和长时间跟踪人体运动的研究分析受到了广泛关注。大多数人体运动都包含了重复的可识别周期,运动中的加速度和速度测量值在一个周期内的特定时期为零值,对周期性步态规律的识别能够帮助我们将跟踪信号中的误差减少,得到更加精确的位置信息。为了使定位的精度不易受到外界的干扰并且不依赖于其他基础建设的支持,惯性传感系统有着完全自主定位的优点,测量单元既不会受到运动模型的影响也不会受到定位环境的影响,因而受到了广泛的关注。然而,受到低成本惯性传感器件本身误差的限制,定位信息并不准确,所以急需要一种计算复杂度低且定位精度高的方案。本文首先简要地介绍和分析了基于捷联式惯性导航中姿态解算算法的研究现状与特点,将毕卡法与有加速度和磁力计融合补偿误差的姿态算法进行性能对比,分析了不同姿态解算针对不同运动模式的优缺点。在此基础上,深入考察了基于梯度下降的姿态更新算法在时间复杂度和性能上的提高,并据此给出了适合人体运动模式下自适应的姿态解算算法。然后提出了将惯性测量单元固定在鞋上的运动跟踪方式,结合步行者运动中传感器信号周期性变化的特点来提高运动轨迹跟踪精度的算法,给出了基于频域的步态周期计算方式以及步态周期中零速度更新算法。最后,在实际环境中,利用差分法提取有效数据片段,采用多采样率处理来自适应运动模式。研究结果表明,经过预处理和多采样率处理后,每个有效运动时间片段内的数据使用于不同的姿态算法,能够得到更加平滑和复杂度低的定位。结合运动模式的特点,将传感器安装在鞋上对步行者运动轨迹跟踪过程中,通过零速度检测进一步降低漂移误差实现对实际环境中运动轨迹的有效跟踪。
[Abstract]:More and more applications need to track and locate people who work indoors and outdoors. In the fields of medicine, games, virtual reality and so on, we need to obtain and track and reconstruct the data of human body movement. Human motion can be described for more potential information. In clinical rehabilitation, tracking patients' movements in outdoor environments allows them to be monitored in real time and corrected in time; in the field of games, a realistic virtual world is created. In the field of interpersonal interaction, intelligent machines not only need to communicate with people directly through voice, but also need more natural interaction methods, such as movement and posture, etc. In more practical fields, people expect to obtain the time of daily movement, the track of walking, running and the law of movement to further adjust the law of their own activities. Therefore, the research and analysis of real-time recording, wireless transmission and long-term tracking of human motion have received wide attention. Most human movements contain repeated identifiable cycles, the acceleration and velocity measurements in motion are zero at a particular period of the cycle, and the identification of periodic gait patterns can help us reduce the errors in tracking signals. Get more accurate location information. In order to make the positioning accuracy difficult to be disturbed by the outside world and not dependent on the support of other infrastructure, the inertial sensing system has the advantage of completely autonomous positioning. The measurement unit is not affected by the motion model or the location environment, so it has been paid more and more attention. However, due to the error of the low cost inertial sensor, the location information is not accurate, so a scheme with low computational complexity and high positioning accuracy is urgently needed. Firstly, this paper briefly introduces and analyzes the research status and characteristics of attitude calculation algorithm based on strapdown inertial navigation, and compares the performance of Bika algorithm with attitude algorithm with acceleration and magnetometer fusion compensation error. The advantages and disadvantages of different attitude solutions for different motion modes are analyzed. On this basis, the improvement of the time complexity and performance of the attitude updating algorithm based on gradient descent is investigated in depth, and an adaptive attitude algorithm suitable for human motion mode is presented. Then, an algorithm is proposed to improve the tracking accuracy of the motion track by fixing the inertial measurement unit on the shoe and combining the characteristics of periodic changes of the sensor signal in the walker motion. The method of calculating gait period in frequency domain and the algorithm of zero velocity updating in gait period are presented. Finally, in the actual environment, the effective data segment is extracted by differential method, and the multi-sampling rate is used to deal with the adaptive motion pattern. The results show that after preprocessing and multi-sampling rate processing, the data in each effective moving time segment can be used in different attitude algorithms, which can achieve smoother and less complexity localization. According to the characteristics of motion mode, sensors are installed on shoes to track the walker's motion track, and the drift error is further reduced by zero velocity detection to realize the effective tracking of the moving track in the actual environment.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN96
本文编号：2154333
[Abstract]:More and more applications need to track and locate people who work indoors and outdoors. In the fields of medicine, games, virtual reality and so on, we need to obtain and track and reconstruct the data of human body movement. Human motion can be described for more potential information. In clinical rehabilitation, tracking patients' movements in outdoor environments allows them to be monitored in real time and corrected in time; in the field of games, a realistic virtual world is created. In the field of interpersonal interaction, intelligent machines not only need to communicate with people directly through voice, but also need more natural interaction methods, such as movement and posture, etc. In more practical fields, people expect to obtain the time of daily movement, the track of walking, running and the law of movement to further adjust the law of their own activities. Therefore, the research and analysis of real-time recording, wireless transmission and long-term tracking of human motion have received wide attention. Most human movements contain repeated identifiable cycles, the acceleration and velocity measurements in motion are zero at a particular period of the cycle, and the identification of periodic gait patterns can help us reduce the errors in tracking signals. Get more accurate location information. In order to make the positioning accuracy difficult to be disturbed by the outside world and not dependent on the support of other infrastructure, the inertial sensing system has the advantage of completely autonomous positioning. The measurement unit is not affected by the motion model or the location environment, so it has been paid more and more attention. However, due to the error of the low cost inertial sensor, the location information is not accurate, so a scheme with low computational complexity and high positioning accuracy is urgently needed. Firstly, this paper briefly introduces and analyzes the research status and characteristics of attitude calculation algorithm based on strapdown inertial navigation, and compares the performance of Bika algorithm with attitude algorithm with acceleration and magnetometer fusion compensation error. The advantages and disadvantages of different attitude solutions for different motion modes are analyzed. On this basis, the improvement of the time complexity and performance of the attitude updating algorithm based on gradient descent is investigated in depth, and an adaptive attitude algorithm suitable for human motion mode is presented. Then, an algorithm is proposed to improve the tracking accuracy of the motion track by fixing the inertial measurement unit on the shoe and combining the characteristics of periodic changes of the sensor signal in the walker motion. The method of calculating gait period in frequency domain and the algorithm of zero velocity updating in gait period are presented. Finally, in the actual environment, the effective data segment is extracted by differential method, and the multi-sampling rate is used to deal with the adaptive motion pattern. The results show that after preprocessing and multi-sampling rate processing, the data in each effective moving time segment can be used in different attitude algorithms, which can achieve smoother and less complexity localization. According to the characteristics of motion mode, sensors are installed on shoes to track the walker's motion track, and the drift error is further reduced by zero velocity detection to realize the effective tracking of the moving track in the actual environment.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN96
【参考文献】
相关期刊论文 前2条
1 罗胤;路鸿洲;赵宇;王君;;基于STM32的步行者航位推算装置设计[J];单片机与嵌入式系统应用;2014年03期
2 肖永健;肖力;孙志刚;;基于步行者航位推算的井下人员辅助定位[J];太赫兹科学与电子信息学报;2013年04期
，本文编号：2154333
本文链接：https://www.wllwen.com/kejilunwen/wltx/2154333.html
