MULTI-BASE STATIONS BLUETOOTH LOCATION METHOD BASED ON FULL CONVOLUTIONAL NEURAL NETWORK

Abstract

Bluetooth positioning system has attracted much attention in the field of indoor navigation due to its high precision, low power consumption and small size. In order to solve the problem that geometric analysis cannot accurately solve the tag position, a multi-base stations Bluetooth location method combined with signal space score map and full convolutional neural network is proposed in the paper. The method generates a signal space score map based on pixels. The signal space score map inputs into the trained full convolutional neural network to output the score map of label position space. The predicted label position is obtained by mapping the pixel point with the maximum value to real spatial coordinates. It's worth mentioning that the actual distance represented by two adjacent pixels is the minimum positioning accuracy. The sigmoid function is used as the activation function in the fully convolutional neural network, so the neural network can represent the predicted uncertainty of tag position in the tag position space score map. The experimental results prove that the system improves the positioning accuracy and the robustness of navigation system.