基于超宽带和支持向量机的人体姿势识别

发布时间：2018-03-14 20:50

本文选题：超宽带　切入点：人体姿势识别　出处：《北京邮电大学》2015年硕士论文　论文类型：学位论文

【摘要】：超宽带(Ultra-wideband, UWB)技术具有多径分辨能力强、穿透能力强以及功耗低的特点,广泛应用于障碍物检测以及目标识别等领域。同时,伴随着人机交互需求的发展,关于人体姿势识别的研究越来越多。本文结合机器学习的理论,提出了挖掘UWB信号传感信息来识别人体姿势的方法。针对基于支持向量机(Support Vector Machine, SVM)的人体姿势识别、改进的混沌自适应遗传算法以及人体姿势识别验证平台三个方面展开了研究,主要工作如下：针对基于SVM的人体姿势识别问题,为了有效提取8种动作UWB信号的传感信息,重点分析了基于小波包分解的特征提取方法。首先,利用小波包分解求出各个频率成分,计算每个频段的能量和,最后得到归一化的小波包能量分布特征。结果表明,小波包能量特征具有良好的可分性,能够显著提高姿势识别的准确率。针对SVM的参数对识别性能影响较大的问题,利用改进的混沌自适应遗传算法对SVM参数进行了优化研究。考虑到标准遗传算法中交叉和变异概率需要预先确定且在算法中维持不变,当整个种群适应度比较接近时,进化将会变慢。本文提出了改进的混沌自适应遗传算法(Improved Chaos Adaptive Genetic Algorithm, ICAGA),它采用动态的交叉和变异概率,且对种群中具有最高适应度的个体进行给定步数的混沌优化搜索,从而指导整个群体向最优解方向进化,改进了遗传算法可能陷入局部最优解的缺陷,并加快了搜索速度。将改进的遗传算法优化应用于人体姿势识别,结果表明,改进算法可以提高SVM的参数寻优速度。在上述分析研究成果的基础上,结合MATLAB的GUI仿真平台,设计开发了基于UWB与SVM的人体姿势识别验证平台,实现了对人体姿势的识别。论文最后对全文工作进行了总结,并对人体姿势识别的相关研究提出了展望。
[Abstract]:Ultra-Wideband Ultra-wideband (UWB) technology is widely used in obstacle detection and target recognition due to its strong multi-path resolution, strong penetration and low power consumption. There are more and more researches on human posture recognition. Based on the theory of machine learning, this paper proposes a method to mine the sensing information of UWB signal to recognize human posture. The improved chaotic adaptive genetic algorithm and the verification platform of human posture recognition are studied. The main work is as follows: aiming at the problem of human posture recognition based on SVM, in order to extract the sensing information of eight kinds of action UWB signals effectively, The feature extraction method based on wavelet packet decomposition is analyzed. Firstly, the energy sum of each frequency band is calculated by wavelet packet decomposition, and the normalized wavelet packet energy distribution is obtained. The energy feature of wavelet packet has good separability and can improve the accuracy of posture recognition significantly. Aiming at the problem that the parameters of SVM have great influence on the recognition performance, The improved chaotic adaptive genetic algorithm is used to optimize the SVM parameters. Considering that the crossover and mutation probabilities in the standard genetic algorithm need to be determined in advance and remain unchanged in the algorithm, when the population fitness is close, This paper presents an improved Chaos Adaptive Genetic algorithm (ICAGAA), which uses dynamic crossover and mutation probability, and performs chaotic optimization search for individuals with the highest fitness in the population. Thus, the whole population is guided to evolve towards the optimal solution, and the defect of the genetic algorithm which may fall into the local optimal solution is improved, and the search speed is accelerated. The improved genetic algorithm is applied to human posture recognition, and the results show that, The improved algorithm can improve the speed of parameter optimization of SVM. Based on the above research results, combined with the GUI simulation platform of MATLAB, a human posture recognition verification platform based on UWB and SVM is designed and developed. Finally, the thesis summarizes the work of the whole paper, and puts forward the prospect of the research on the recognition of human posture.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN925;TP181

【相似文献】