可穿戴计算中基于情景感知的能效控制策略研究

发布时间：2018-06-11 23:05

本文选题：可穿戴计算 + 惯性测量单元　；参考：《南京邮电大学》2017年硕士论文

【摘要】：近年来,随着微控制单元(Micro Control Unit,MCU)、微电子机械系统(Microelectromechanical Systems,MEMS)和无线传感网(Wireless Sensor Network,WSN)的快速发展,基于无线可穿戴设备的动作捕捉系统具有良好的应用前景和巨大的商业价值。惯性测量单元(Inertial Measurement Unit,IMU)具有体积小、便于携带以及成本低等优点,目前在医疗康复、老人监护等重要领域已得到广泛应用。然而由于需要长时间运行传感节点和高频高速率实时传输数据流等,基于惯性测量单元的动作捕捉系统存在能效低、续航能力弱等问题。针对上述问题,本文研究在低功耗约束环境下,结合情景感知技术,探索可穿戴计算中基于惯性测量单元的动作捕捉系统的能效提高策略,并在无线可穿戴平台Shimmer2R上对准确率优化和能耗降低方面进行了验证。主要研究成果如下:(1)实现了腕戴式单传感器动作捕捉系统。目前较为流行的动作捕捉系统采用的均是三轴加速度传感器,现有的系统存在有线牵制、佩戴节点多、佩戴设备重等不足。针对当前存在的问题,本文提出一种腕戴式运动识别系统,采用佩戴在实验者右手手腕上的单个惯性传感器进行数据采集。在不影响使用者日常活动的前提下满足佩戴的舒适度,并且能够达到较好的行为识别准确率,同时降低系统的计算复杂度。(2)实现了人体运动识别优化算法,用于提高动作捕捉系统的能效。本文设计了适用于人体运动识别分类模型的优化算法,用于提高运动识别系统的整体能效性。以往的算法中由于分类模型存在参数选定问题,默认的参数值不能适应不同个体的行动特点。本文提出对于不同的分类模型参数分别进行优化,针对不同个体的运动习惯求得最佳参数组合,达到提升人体运动识别的分类准确率。实验结果表明,系统优化算法能够有效提高人体分类准确率,分类准确率最高可以达到98.59%。相比传统分类器,分类准确率最高可以提高26.05%。(3)提出了自适应无线传输控制算法和多分类器融合的自适应控制算法。动作捕捉系统所采用的无线可穿戴节点的能耗主要集中在无线传输通信模块。高速实时的无线传输消耗了传感节点大部分的能量。本文提出自适应无线传输控制算法,结合情景感知技术,对人体运动数据进行处理判断,并能自适应控制无线传输的数据量,可以有效的降低无线传输的能耗。并在此基础上,提出了多分类器融合的自适应数据传输控制。实验结果表明,提出的自适应数据传输控制方法能够在不降低分类准确率的前提下,减少无线通信的数据传输量,数据传输削减量最高可达76.7%。通过此方法提高设备能效性和节点的续航能力,续航时间最高可延长73.38%。
[Abstract]:In recent years, with the rapid development of Micro Control Unit (MCU), Microelectromechanical Systems (MEMSs) and Wireless Sensor Network (WSNs), the motion capture system based on wireless wearable devices has a good application prospect and great commercial value. Inertial Measurement Unit (IMU) has many advantages such as small size, easy to carry and low cost. At present, it has been widely used in many important fields, such as medical rehabilitation, care for the elderly and so on. However, due to the need of long time operation of sensor nodes and real-time transmission of high frequency and high speed data streams, the motion capture system based on inertial measurement unit has some problems, such as low energy efficiency and weak endurance. In order to solve the above problems, this paper studies the strategy of improving the energy efficiency of the motion capture system based on inertial measurement unit in wearable computing under low power constraints and scenario sensing technology. The accuracy optimization and energy consumption reduction are verified on Shimmer2R, a wireless wearable platform. The main research results are as follows: 1) the single sensor motion capture system is realized. At present, the popular motion capture system uses three-axis acceleration sensor, the existing system has some shortcomings, such as cable traction, wearing more nodes, wearing equipment weight, and so on. Aiming at the existing problems, this paper presents a motion recognition system based on wrist wearing, which uses a single inertial sensor which is worn on the right hand wrist of the experimenter to collect data. On the premise of not affecting the user's daily activities, it can satisfy the comfortable degree of wearing, and can achieve better accuracy of behavior recognition, and reduce the computational complexity of the system. At the same time, it realizes the optimization algorithm of human motion recognition. Used to improve the energy efficiency of motion capture systems. In this paper, an optimization algorithm suitable for classification model of human motion recognition is designed to improve the overall energy efficiency of motion recognition system. In the previous algorithms, because of the problem of parameter selection in the classification model, the default parameter values can not adapt to the behavior characteristics of different individuals. In this paper, the parameters of different classification models are optimized, and the best combination of parameters is obtained according to different individuals' movement habits, so as to improve the classification accuracy of human motion recognition. Experimental results show that the system optimization algorithm can effectively improve the accuracy of human classification, the highest classification accuracy can reach 98.5959. Compared with the traditional classifier, the classification accuracy can be improved by 26.05. (3) the adaptive wireless transmission control algorithm and the multi-classifier fusion adaptive control algorithm are proposed. The energy consumption of the wireless wearable nodes used in the motion capture system is mainly concentrated in the wireless transmission communication module. High-speed and real-time wireless transmission consumes most of the energy of sensor nodes. In this paper, an adaptive wireless transmission control algorithm is proposed, which can process and judge human motion data with scene sensing technology, and can adaptively control the amount of wireless transmission data, which can effectively reduce the energy consumption of wireless transmission. On this basis, a multi-classifier fusion adaptive data transmission control is proposed. The experimental results show that the proposed adaptive data transmission control method can reduce the amount of wireless communication data transmission without reducing the classification accuracy, and the maximum reduction amount of data transmission can be up to 76.7. By using this method, the energy efficiency of the equipment and the capacity of the node can be improved, and the maximum life span can be extended by 73.38.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP368.33

【参考文献】