基于标签信息的全局流形对齐定位算法

Global manifold alignment localization algorithm based on label information

随着WLAN的广泛部署及应用,基于位置指纹的室内定位技术成为研究的热点。针对该技术需要密集布设并采集参考点接收的RSS信号,完成位置标记而带来指纹数据构建繁琐的问题,提出利用部分标签数据,结合实时采集未标记数据,实现多域间高维数据的低维嵌入。在提取低维特征的同时,保持流形的全局结构相似性不变,实现对未标记数据的位置解算。实验结果表明,在同一定位环境下,本算法具有较好的定位精度,能够满足目前对室内定位的需求。

With the wide deployment and application of wireless local area network(WLAN),location fingerprint technology is a hot topic in the field of indoor location.The construction of intensive reference points layout and collecting RSS signals received,obtaining locations is time and labor consuming.In order to solve this problem,we propose an approach embedding multi-domain high-dimensional data into low-dimensional.The proposed method utilizes part of labeled RSS data,combines with unlabeled data collected in real time.Then,the locations of unlabeled data can be obtained by extracting low-dimensional features on the condition of keeping the global structural similarity invariant of manifolds.The experimental results show that the proposed algorithm can achieve high localization accuracy in the same positioning environment which can meet the needs of indoor positioning requirements.