A Robust WiFi Localization Algorithm Using Data Augmentation and Stacked Denoising Autoencoder

WiFi-based fingerprint localization technology has become one of the most practical methods for localizing mobile users due to its non-intrusive nature, low cost and no additional equipment required. However, the fluctuation of WiFi signal seriously affects the accuracy of WiFi fingerprint localization. To address this problem, this paper proposes a solution using data augmentation combined with stacked denoising autoencoder (SDAE). Data augmentation can facilitate the neural network to learn the mapping relationship between the fluctuating WiFi signals and coordinates. And the SDAE can obtain a robust and time-independent feature from the dynamic WiFi signal. A convolutional neural network is also used to build a floor classification model to determine the height, and a multilayer perceptron (MLP) is used to build a regression model to determine the relative coordinates. Experimental results on public datasets show that the method improves system robustness and localization accuracy.