POSTER: Wi-Fi Indoor Positioning Based on Sparse Autoencoder and Deep Belief Network

Abstract

In order to reduce the influence of the complexity and diversity of indoor environment on traditional localization methods and to more effectively use Wi-Fi fingerprint data to position an object, an indoor localization algorithm based on Sparse autoencoder(SAE) and Deep Belief Network(DBN) was proposed, the SAE-DBN model, was proposed. In this method, the SAE first extracts the depth features of the training data, and identifies the objects from different experimental areas. Then, the DBN model of the corresponding area is used to accurately position the objects. The simulation results show that compared with the traditional Wi-Fi positioning method and some existing improved algorithms, the proposed Wi-Fi positioning method has higher accuracy and stability, and the average positioning accuracy is 1.13 m.