Two indoor location algorithms based on sparse fingerprint library

The positioning scheme which based on the RSSI fingerprint database has started to spring up in recent years. However, due to the instability of the RSSI signal, the effect of the fingerprint database positioning scheme is not remarkable. In order to obtain high-precision positioning results, we require intensive fingerprint data which is difficult to sample. Therefore, we propose a scheme of fitting a dense fingerprinting database through the sparse fingerprint database, and adopt two improved machine learning algorithms (WKNN and RBF neural network) to establish the location model Experiments show that it is reliable to build dense fingerprint database through measuring sparse fingerprints firstly and use WKNN or RBF neural network to establish a localization model. At last we achieved the positioning accuracy less than 25 cm within a range of 5 m × 5 m