A novel radio-map establishment based on fuzzy clustering for hybrid K-Nearest Neighbor (KNN) and Artifi cial Neural Network (ANN) position algorithm in WLAN indoor environment is proposed. First of all, the Principal...A novel radio-map establishment based on fuzzy clustering for hybrid K-Nearest Neighbor (KNN) and Artifi cial Neural Network (ANN) position algorithm in WLAN indoor environment is proposed. First of all, the Principal Component Analysis (PCA) is utilized for the purpose of simplifying input dimensions of position estimation algorithm and saving storage cost for the establishment of radio-map. Then, reference points (RPs) calibrated in the off-line phase are divided into separate clusters by Fuzzy C-means clustering (FCM), and membership degrees (MDs) for different clusters are also allocated to each RPs. However, the singular RPs cased by the multi-path effect signifi cantly decreases the clustering performance. Therefore, a novel radio-map establishment method is presented based on the modifi cation of signal samples recorded at singular RPs by surface fitting. In the on-line phase, the region which the mobile terminal (MT) belongs to is estimated according to the MDs firstly. Then, in estimated small dimensional regions, MT's coordinates are calculated byKNN positioning method for efficiency purpose. However, for the regions including singular RPs, ANN method is utilized because ofits great pattern matching ability. Furthermore, compared with other typical indoor positioning methods, feasibility and effectiveness of this hybrid KNN/ANN method are also verified by the experimental results in static and tracking situations.展开更多
WiFi室内定位已被广泛研究,并且提出了许多解决方案,其中以接收信号强度(received signal strength,RSS)作为位置指纹的加权K-最近邻(weighted K-Nearest neighbor,WKNN)算法是目前使用最广泛的位置指纹算法之一。由于WKNN算法通常采用...WiFi室内定位已被广泛研究,并且提出了许多解决方案,其中以接收信号强度(received signal strength,RSS)作为位置指纹的加权K-最近邻(weighted K-Nearest neighbor,WKNN)算法是目前使用最广泛的位置指纹算法之一。由于WKNN算法通常采用固定的K值,其定位精度在实际使用时具有局限性。尽管动态K的方案被提出,但是由于引入了新的不确定性参数,因此,并未真正解决问题。针对这个问题,提出了一种自适应动态K的WKNN室内定位方法。提出的算法的K值自适应调整仅依赖于离线和在线数据,即可以不引入新的不确定参数。在这个前提下,提出的算法采用"多雷达搜索策略"的方式自适应选择近邻数K值进行在线位置估计。在真实环境中采样了大量数据进行了试验。试验结果表明,提出的算法可根据在线情况自适应调整K值,获得了较好的定位结果。展开更多
基金supported by National High-Tech Research & Development Program of China (Grant No. 2008AA12Z305)
文摘A novel radio-map establishment based on fuzzy clustering for hybrid K-Nearest Neighbor (KNN) and Artifi cial Neural Network (ANN) position algorithm in WLAN indoor environment is proposed. First of all, the Principal Component Analysis (PCA) is utilized for the purpose of simplifying input dimensions of position estimation algorithm and saving storage cost for the establishment of radio-map. Then, reference points (RPs) calibrated in the off-line phase are divided into separate clusters by Fuzzy C-means clustering (FCM), and membership degrees (MDs) for different clusters are also allocated to each RPs. However, the singular RPs cased by the multi-path effect signifi cantly decreases the clustering performance. Therefore, a novel radio-map establishment method is presented based on the modifi cation of signal samples recorded at singular RPs by surface fitting. In the on-line phase, the region which the mobile terminal (MT) belongs to is estimated according to the MDs firstly. Then, in estimated small dimensional regions, MT's coordinates are calculated byKNN positioning method for efficiency purpose. However, for the regions including singular RPs, ANN method is utilized because ofits great pattern matching ability. Furthermore, compared with other typical indoor positioning methods, feasibility and effectiveness of this hybrid KNN/ANN method are also verified by the experimental results in static and tracking situations.
文摘WiFi室内定位已被广泛研究,并且提出了许多解决方案,其中以接收信号强度(received signal strength,RSS)作为位置指纹的加权K-最近邻(weighted K-Nearest neighbor,WKNN)算法是目前使用最广泛的位置指纹算法之一。由于WKNN算法通常采用固定的K值,其定位精度在实际使用时具有局限性。尽管动态K的方案被提出,但是由于引入了新的不确定性参数,因此,并未真正解决问题。针对这个问题,提出了一种自适应动态K的WKNN室内定位方法。提出的算法的K值自适应调整仅依赖于离线和在线数据,即可以不引入新的不确定参数。在这个前提下,提出的算法采用"多雷达搜索策略"的方式自适应选择近邻数K值进行在线位置估计。在真实环境中采样了大量数据进行了试验。试验结果表明,提出的算法可根据在线情况自适应调整K值,获得了较好的定位结果。
