WLAN Indoor Positioning Based on D-LDA Feature Extraction Algorithm

This paper introduces the Direct Linear Discriminant Analysis (D-LDA) algorithm for feature extraction to reduce noise and redundant location information of the access points (APs) signals in wireless LAN (WLAN) indoor positioning system. Feature database is obtained by deploying D-LDA algorithm to extract the low-dimensional and discriminative positioning features from the original WLAN signal database. The dimensionality of the extracted features may be chosen by setting appropriate retained eigenvalues ratio of between-class scatter matrix. Based on the generated feature database, three typical localization algorithms including weighted k-nearest neighbor (WKNN), nearest-neighbor (NN) and maximum likelihood (ML) are carried for real-time positioning and the results are compared. D-LDA feature extraction algorithm obtains the higher accuracy than traditional localization algorithms while reducing the storage and computation cost significantly.