Multi-bandwidth NLOS Identification Based on Deep Learning Method
Tiantian Chang, Suying Jiang, Yuzhe Sun, Ailin Jia, Wei Wang
Abstract
The identification of the propagation conditions of the wireless system, i.e., line-of-sight (LOS) and non-line-of-sight (NLOS), has a significant impact on radio based positioning performance and channel modeling. Traditional methods to identify LOS/NLOS are usually based on extracting features of the channel impulse response (CIR), based on which a decision threshold is used to distinguish between LOS and NLOS conditions. However, channel features of LOS and NLOS are sometimes similar with each other, which usually results in low accuracy of LOS/NLOS identification. In this paper, we use the combined channel feature that is consist of channel state information (CSI) and four characteristics of the channel impulse response as input, and use Long Short-Term Memory (LSTM) to train the learning method to identify LOS/NLOS condition. The proposed method is evaluated based on measured data. The results show that the proposed method achieves an accuracy of 95.477% for LOS/NLOS channel identification.