基于加速度传感器的人体姿态识别

发布时间：2018-05-16 02:17

本文选题：加速度传感器 + 人体姿态识别　；参考：《长沙理工大学》2016年硕士论文

【摘要】：近些年来,随着微型机电系统(MEMS)技术的飞速发展和模式识别理论研究的不断深入,基于加速度传感器的人体姿态识别已经成为了人体姿态识别领域中一个重要的研究方向,在运动分析、医疗监护、体感游戏和能耗评估等领域受到了广泛的关注。相比于基于图像分析的人体姿态识别,该方法不受环境约束,成本低,能耗少,拥有更为广阔的应用前景。当前,基于加速度传感器的人体姿态识别研究仍处在一个比较基础的阶段,由于客观环境的多样性和人体运动的复杂性使得基于加速度传感器的人体姿态识别研究还存在很多急需解决的问题,其中包括如何提取具有更强表征能力的数据特征,如何设计高效、精确的分类器等。围绕这些问题和难点,本文针对基于加速度传感器的人体姿态识别技术展开了一系列的研究,主要的工作如下:(1)、总结了现有的人体姿态识别方法,比较了基于图像分析和基于加速度传感器的两种方法,系统地对数据采集、数据预处理、特征提取和选取、分类算法等模块进行了分析。(2)、提出了一种基于改进粒子群(PSO)优化神经网络的人体姿态识别算法。首先在常用特征集的基础上,加入离散系数和曲线积分这两种能够反映加速度变化趋势和速度变化量的特征作为神经网络的输入;然后在利用PSO优化神经网络参数的同时,通过控制概率,自适应地对粒子进行遗传操作,增强了粒子跳出局部极小值的能力,并利用训练得到的分类模型对6种人体姿态进行识别,实验结果发现改进PSO优化神经网络的收敛速度和全局寻优能力得到了提高,与其他经典分类算法相比识别精度更高。(3)、提出了一种基于窗口相似度的人体姿态识别算法。首先采用曲线拟合对离散加速度数据进行处理,并利用粒子群算法优化关键参数后得到其拟合曲线;然后计算不同拟合曲线之间的窗口相似度,并以窗口相似度作为距离度量,采用K近邻分类算法来识别人体姿态。通过实验对比说明该算法的计算量较小,且能准确地识别6种人体姿态。总而言之,基于加速度传感器的人体姿态识别仍处在发展阶段,该课题的研究具有很大的理论价值和实际需要,值得人们去进行更深入、细致的研究。
[Abstract]:In recent years, with the rapid development of MEMS technology and the development of pattern recognition theory, human attitude recognition based on acceleration sensor has become an important research direction in the field of human attitude recognition. Sports analysis, medical monitoring, somatosensory games and energy consumption evaluation have received extensive attention. Compared with the human body attitude recognition based on image analysis, this method is not subject to environmental constraints, low cost, less energy consumption, and has a wider application prospect. At present, the research of human body attitude recognition based on acceleration sensor is still in a relatively basic stage. Due to the diversity of objective environment and the complexity of human motion, there are still many problems that need to be solved in the research of human attitude recognition based on acceleration sensor, including how to extract the data features with stronger representation ability. How to design efficient and accurate classifiers. Focusing on these problems and difficulties, this paper has carried out a series of research on the technology of human body attitude recognition based on acceleration sensor. The main work is as follows: 1. The existing methods of human body attitude recognition are summarized. Two methods based on image analysis and acceleration sensor are compared. Data acquisition, data preprocessing, feature extraction and selection are systematically compared. The classification algorithm and other modules are analyzed, and a human body attitude recognition algorithm based on improved particle swarm optimization (PSO) neural network is proposed. Firstly, the discrete coefficient and curve integral, which can reflect the trend of acceleration and the variation of velocity, are added as the input of neural network on the basis of common feature sets, and then the parameters of neural network are optimized by using PSO. By controlling probability and adaptively genetic manipulation of particles, the ability of particles to jump out of local minimum is enhanced, and six kinds of human posture are recognized by the trained classification model. The experimental results show that the convergence speed and global optimization ability of the improved PSO optimization neural network are improved and the recognition accuracy is higher than other classical classification algorithms. A human body attitude recognition algorithm based on window similarity is proposed. Firstly, the discrete acceleration data are processed by curve fitting, and the key parameters are optimized by particle swarm optimization algorithm to get the fitting curve, then the window similarity between different fitting curves is calculated, and the window similarity is taken as the distance measure. K nearest neighbor classification algorithm is used to recognize human posture. The experimental results show that the algorithm is less computational and can accurately identify six human body postures. In a word, the attitude recognition of human body based on acceleration sensor is still in the development stage, the research of this subject has great theoretical value and practical need, and it is worth people to carry out more in-depth and detailed research.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TP212

【参考文献】