基于CSI的室内指纹定位算法研究

发布时间：2018-01-11 00:29

本文关键词：基于CSI的室内指纹定位算法研究　出处：《太原理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着Wi-Fi技术的不断成熟以及无线设备在室内的广泛部署,基于无线信号的室内定位技术受到了越来越多的关注。其中基于位置指纹的定位技术由于其成本低、应用范围广且无需额外硬件支持等优势,成为了室内定位领域研究的重点。近年来,由于硬件设备不断支持物理层信道状态信息的获取,基于CSI(Channel State Information)的指纹定位技术为室内定位的研究提供了新思路。由于CSI提供了更精细的信道状态信息,而且受多径效应影响小,信号特征变化较RSSI更加稳定,因此CSI的出现为室内指纹定位技术的研究与发展引领了新的方向。目前大多基于CSI的指纹定位系统在采样阶段只利用信道的振幅信息作为位置的指纹特征,对信道状态的分析不够完善;而且在定位阶段通常不考虑定位环境相对于基准环境的变化,仅利用现有的算法进行指纹特征的匹配,致使定位误差较大,定位精度仍需提高。本文针对上述指纹定位系统存在的不足,通过分析信道的CSI特征,利用数学方法提取到了可靠的相位信息,增加了定位时可用的信息量。同时,针对不同的室内环境,提出了两种定位解决方案:首先,在室内环境信道稳定的情况下,提出了基于相关系数的指纹定位算法。该算法通过验证节点间相关系数与位置之间的关系,利用矩阵间的相关系数来获取两个地点或两种状态间的匹配概率,最后利用匹配概率最大和次大的点的坐标来估算待测点的最终位置。与Bayes算法相比,该方法降低了概率计算的复杂度,实现了更便捷的位置估计。其次,当定位环境相对于采样环境发生较大的变化时,提出了自适应修正模型定位算法。在信道不稳定的情况下,该算法通过引入一个基于采样环境的扰动因子PEM(Percentage of nonzero Elements),来反映定位环境相对于基准环境的变化,然后利用建立的自适应修正模型完成指纹数据库的更新;最后通过改进的AWKNN(AP weighted and distanced weighted K Nearest Neighbor)算法实现了待测点的位置估计。实验结果证明:定位阶段信道稳定时,基于相关系数的定位算法通过引入相位信息,使待测点的平均定位误差从1.18m降低到了0.75m,而且相比于Bayes算法,相关系数算法大大降低了定位计算的复杂度,提高了系统的定位性能。在信道不稳定时,自适应修正模型定位系统相比之前的指纹定位系统FIFS和CSI-MIMO,准确率分别提高了67%和25%,而且自适应修正模型算法实现了不同环境下系统的自适应定位,进一步推动了室内定位技术的发展。
[Abstract]:With the continuous maturity of Wi-Fi technology and the extensive deployment of wireless devices in the room. The indoor location technology based on wireless signal has been paid more and more attention, among which the location fingerprint based location technology has the advantages of low cost, wide application range and no additional hardware support. In recent years, because of the hardware equipment support the physical layer channel state information acquisition. Based on CSI(Channel State Information). The fingerprint location technology based on CSI provides a new idea for indoor localization research, because CSI provides more detailed channel state information. Moreover, it is less affected by multipath effect, and the change of signal characteristics is more stable than that of RSSI. Therefore, the emergence of CSI has led a new direction for the research and development of indoor fingerprint location technology. At present, most fingerprint localization systems based on CSI only use the amplitude information of the channel as the fingerprint feature in the sampling stage. Sign. The analysis of channel state is not perfect; In the localization stage, the change of the location environment relative to the reference environment is not considered, only the existing algorithms are used to match the fingerprint features, resulting in a large positioning error. The location accuracy still needs to be improved. Aiming at the shortcomings of the fingerprint location system mentioned above, the reliable phase information is extracted by mathematical method by analyzing the CSI characteristics of the channel. At the same time, for different indoor environment, two localization solutions are proposed: first, in the case of indoor environment channel stability. A fingerprint location algorithm based on correlation coefficient is proposed, which verifies the relationship between the correlation coefficient and the location between nodes, and obtains the matching probability between two locations or two states by using the correlation coefficients between matrices. Finally, the coordinates of the points with the maximum and sub-maximum matching probability are used to estimate the final position of the points to be measured. Compared with the Bayes algorithm, this method reduces the complexity of the probability calculation. A more convenient location estimation is realized. Secondly, when the location environment changes compared with the sampling environment, an adaptive modified model location algorithm is proposed. In the case of unstable channel, an adaptive modified model location algorithm is proposed. The algorithm introduces a PEM(Percentage of nonzero elements based on sampling environment. To reflect the change of the location environment relative to the reference environment, and then use the self-adaptive correction model to complete the fingerprint database update; Finally, the improved AWKNN(AP weighted and distanced weighted K Nearest neighbor was adopted. The experimental results show that the channel is stable in the positioning phase. The algorithm based on correlation coefficient can reduce the average positioning error from 1.18m to 0.75m by introducing phase information, and compared with Bayes algorithm. Correlation coefficient algorithm greatly reduces the complexity of location computation and improves the positioning performance of the system. Compared with FIFS and CSI-MIMO-based fingerprint location system, the accuracy of adaptive modified model localization system is improved by 67% and 25%, respectively. The adaptive modified model algorithm realizes the adaptive location of the system in different environments, which further promotes the development of indoor positioning technology.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN92

【相似文献】