CSI-based Fingerprinting for Indoor Localization with Multi-scale Convolutional Neural Network

With the rapid-growing demand for location-based services in indoor environments, fingerprint-based indoor positioning methods have attracted great interest owing to high accuracy and low online complexity. In this paper, we use the channel state information (CSI) of the massive MIMO (MaMIMO) system as the fingerprint to construct the fingerprint database. Different from the previous methods that only use CSI amplitude to construct fingerprints, phase information and angle of arrival (AOA) are added to the fingerprint to enhance the characteristics of fingerprint data. We modified the network according to the characteristics of fingerprint data based on Google Net and implemented a GoogleNet-like convolutional neural network(CNN) which uses skip connection and 1-D convolution kernel for fingerprint positioning. Experiment results show that with sufficient representative data sets, centimeter-level positioning can be achieved by using the proposed neural network, and the positioning accuracy can be further improved by 10% with the use of AOA information.