Efficient indoor propagation channel prediction based on deep learning approach

Abstract

In this study, efficient channel characterisation and modelling based on deep learning algorithms are developed and presented in a line-of-sight (LOS) scenario. The learning and the validation processes are performed using measurements from only one environment, enabling robust model learning and prediction results. Then, the model efficiency is analysed and validated using measurements from different environments that are not included in the learning process. Finally, the channel characterisation is made with the predicted and measured ones. The designed model achieved a highly accurate channel frequency response prediction within different environments without any prior information. The model root-mean-square error achieves up to 2% compared with the latest proposed models in the literature. Hence, an efficient modelling tool is provided for the future wireless communication design in a complex confined environment in LOS scenarios.