WiFi indoor location determination via ANFIS with PCA methods

Abstract

This paper proposes the WiFi indoor location determination method based on adaptive neuro-fuzzy inference system (ANFIS) with principal component analysis (PCA). It reduces the WiFi signal vectors dimensions and saves the storage cost and simplifies the fuzzy rules generated by subtractive clustering method for ANFIS training. In the off-line phase, the received signal strength (RSS) or signal to noise ratio (SNR) from multiple access points (APs) is recorded for the establishment of radio map. And in the on-line phase, two steps should be considered for the position determination. The first step is space transformation to principal component space with lower dimensions compared to original space for the signal vectors. And the second step is the estimation of real two or three dimensional coordinates of mobile terminal (MT). Feasibility and effectiveness of ANFIS system based on FCA method are verified according to the analysis of the iterative number of training and experimental comparison with K-nearest neighbor (KNN), probability, artificial neural network (ANN) and ANFIS indoor location system without FCA.