基于半监督学习的室内WLAN支持向量回归定位算法

发布时间：2018-02-15 13:05

本文关键词： WLAN 指纹定位法支持量向量回归半监督学习协同训练　出处：《重庆邮电大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着移动智能终端设备的普及和通信技术的快速发展,基于位置的服务的市场需求越来越大。基于位置的服务在导航、紧急救援、个性化信息的传递等领域发挥着巨大的作用。对于室外定位技术,主要有以美国的GPS为代表的卫星定位技术和利用通信基站的蜂窝网定位技术。然而由于室内环境复杂,室外定位技术在室内很难满足人们对定位精度的要求。同时,随着WLAN设备在室内各种环境中被广泛部署,这为WLAN定位技术的发展和推广奠定了很好的基础。基于WLAN的室内定位技术因其较高的定位精度和不需要额外的设备等优点成为研究的热点。基于位置指纹的WLAN定位技术是WLAN定位技术的主流,其分为离线阶段和在线阶段。本论文正是对基于位置指纹的WLAN定位算法进行研究。首先,由于室内环境复杂,以及无线信号的传播特性,室内接收信号具有不确定性和非线性特性。这都对基于位置指纹的WLAN室内技术的定位性能产生了很大的影响。同时,支持向量机在解决小样本和非线性问题上有很大的优势,且具有很好的泛化能力。基于此,本文把支持向量回归引入到室内WLAN指纹定位中,建立RSS信号与物理位置的映射预测模型,以提高定位精度。其次,由于指纹定位在离线阶段需要花费大量人力物力采集大量的位置指纹,而独立于位置的未标记RSS通过移动终端很容易获得,半监督学习能够很好的利用独立于位置的RSS信息,减少了对位置指纹的要求,同时能够提高定位精度。因此,本文引入半监督学习协同训练算法与支持向量回归相结合,提出基于半监督学习的室内WLAN支持向量回归定位算法,提高定位精度。最后,对基于半监督学习的室内WLAN支持向量回归定位算法进行改进,改善其性能。本文在仿真环境和普遍真实室内环境——办公环境和走廊环境下,对本文提出的算法进行仿真及实验验证。通过与传统算法在性能上的对比,验证本文提出的定位算法在定位性能上的优越性。
[Abstract]:With the popularization of mobile intelligent terminal devices and the rapid development of communication technology, the market demand for location-based services is increasing. The field of personalized information transmission plays a great role. For outdoor positioning technology, there are mainly satellite positioning technology represented by GPS of the United States and cellular network positioning technology using communication base stations. However, because of the complexity of indoor environment, Outdoor positioning technology is very difficult to meet the requirements of positioning accuracy in indoor. At the same time, with the wide deployment of WLAN equipment in various indoor environments, This has laid a good foundation for the development and popularization of WLAN positioning technology. The indoor positioning technology based on WLAN has become a hot spot for its high positioning accuracy and no need of additional equipment. WLAN location based on position fingerprint has become a hot topic. Bit technology is the mainstream of WLAN positioning technology, It is divided into offline phase and online stage. This thesis is to study the location fingerprint based WLAN localization algorithm. Firstly, because of the complexity of indoor environment and the propagation characteristics of wireless signal, Indoor received signals are uncertain and nonlinear, which have great influence on the localization performance of WLAN indoor technology based on position fingerprint. At the same time, support vector machine has great advantages in solving small samples and nonlinear problems. Based on this, support vector regression is introduced into indoor WLAN fingerprint location, and the mapping and prediction model of RSS signal and physical position is established to improve the location accuracy. Because fingerprint location requires a lot of manpower and material resources to collect a large number of location fingerprints, and the location independent RSS can be easily obtained through mobile terminals, semi-supervised learning can make good use of location-independent RSS information. The requirement of location fingerprint is reduced and the location accuracy is improved. Therefore, a semi-supervised learning cooperative training algorithm is combined with support vector regression, and an indoor WLAN support vector regression location algorithm based on semi-supervised learning is proposed in this paper. Finally, the indoor WLAN support vector regression algorithm based on semi-supervised learning is improved to improve its performance. The performance of the proposed algorithm is compared with that of the traditional algorithm, and the superiority of the proposed algorithm in location performance is verified.
【学位授予单位】：重庆邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN925.93

【参考文献】