WLAN indoor localization method using angle estimation

In WLAN indoor localization systems, an improved position fingerprinting algorithm is proposed to obtain higher accuracy. The algorithm constructs the nonlinear relationship between received signal strength indication (RSSI) values and the angles formed by horizontal line and the line from transmitters to receivers, instead of traditionally training the relationship between RSSI values and physical coordinates. The localization area is divided into a number of small rectangular areas, and the test points are sorted out by K-Nearest-Neighbor (KNN) algorithm. In a small rectangular area, RSSI values and the angles are trained by support vector machine (SVM), so as to estimate the angles formed by horizontal line and the line from test points to each access point (AP). Finally, coordinates of the test points are estimated using the geometric relationship. Two experimental sections have been conducted under different conditions: one is in the laboratory, and the other in a typical office space. The proposed algorithm is compared with v-SVM algorithm, KNN algorithm and ML algorithm. Experimental results prove that our proposed algorithm outperforms other methods in term of localization accuracy under various situations.