Deep Neural Network Based on Feature Fusion for Indoor Wireless Localization

This paper proposes a novel WiFi-based indoor localization algorithm based on feature fusion based deep neural network, which is called FF-DNN. The FF-DNN model architecture consists of two phases, off-line training and on-line localization. In off-line training phase, RSS and their corresponding ranking together make up the DNN features, and FF-DNN weights are trained using the mean square error as a loss function. In the on-line localization phase, the target location will be estimated by which the RSS values of each access point (AP) are collected. Experimental results show that, compared with other existing indoor positioning methods, the proposed algorithm can effectively improve the positioning accuracy.