Intelligent Index Classification Method Based on Machine Learning for Detection of Reference Signal in 5G Networks

In order to maintain stable communication in 5G wireless networks, the link between a 5G base station and user equipment (UE) should be constantly monitored and adapted to the time-varying wireless channel. The use of UE for seamless information exchange is based on obtaining a target reference signal. The method used to obtain the reference signal involves identifying the index of the reference signal received from the 5G base stations. However, the existing index identification method employed in commercial 5G networks is based on the blind detection method, which is inefficient in terms of time and can cause misdetections. On the other hand, machine learning (ML), which is statistically predictable through data accumulation, can be robust in practical network environments. Taking this into account, we build a dataset consisting of reference signal data collected in a real-world 5G network environment to obtain an optimal machine learning model that is applicable to practical 5G networks. We evaluate a total of 23 index classification models by combining six ML models and three data pre-processing methods. The results of the study represent optimized combinations of ML-based index classifiers and data pre-processing methods. Performance differences between neural network (NN) models and non-NN models are also revealed.