基于SALDE-UKF-SVM算法的WLAN室内定位方法

摘要

针对室内复杂环境,WLAN信号强度信息高维时变特性,提出一种引入监督能力的自适应局部线性判别嵌入算法(SALDE)和改进支持向量机(SVM)的室内无线定位算法.首先,该算法利用SALDE对所采集的WLAN信号进行特征提取,达到降低维度和增大类别间判别信息的双重作用.然后,在低维流形空间中,利用SVM对数据进行特征分类判别,缩小定位区域,同时建立位置坐标与信息强度的非线性映射模型;最终利用无迹卡尔曼滤波算法(UKF)对估算位置进行滤波处理,提高定位精度与稳定性.仿真结果表明,该算法在定位误差2 m范围内精度达到72.4％,在4 m范围内精度已经高达95.8％,相比于传统SVM算法2m内精度提高18.2％,在4m内的精度提高17.7％,定位精度得到明显提升,可以较好地满足室内定位的需求.%In order to overcome the problems of the complicated indoor environment and deal with the high dimensions and time-varying characteristics of WLAN received signal strength indication (RSSI), an indoor location method based on supervised adaptive local discriminant embedding (SALDE) and UKF-SVM is proposed. First, the SALDE algorithm is employed to extract the features of the collected RSSI and achieve the dual effect of reducing the dimension and increasing the discrimination between categories. Then, in the low-dimensional manifold space, the SVM is used to classify these data, shrink down the positioning area, and establish the nonlinear mapping model of position coordinates. Finally, the positioning result is filtered by using the unscented Kalman filter algorithm. Simulation results show that, compared with the traditional SVM methods, the accuracies of the algorithm within 2m and 4m of positioning errors achieve 72.4％ and 95.8％ respectively, which are improved by 18.2％ and 17.7％ respectively. The proposed method can signify-cantly improve the positioning accuracy and can better meet the accuracy requirements of indoor positioning.