Using Generative Adversarial Nets to Reduce Fingerprint Collection for Indoor Localization

Abstract

WiFi positioning is currently the more mainstream indoor positioning method, and fingerprint database construction is crucial to WiFi-based localization systems. However, this approach requires enough fingerprint data for a single point. In this paper, we convert channel state information (CSI) data into amplitude feature maps to construct initial fingerprint library and then extend the fingerprint database using the proposed improved deep convolutional generative adversarial nets (IDCGAN) model. Finally, the amplitude feature maps are trained by the CNN to locate. Based on the extended fingerprint database, the accuracy of indoor localization systems can be improved with reduced human effort.