Prediction of Signal Quality and SFN Interference Metrics Using Machine Learning Models

Abstract

This paper addresses a machine learning-based approach to the study of the effect of interference in single frequency networks (SFNs). The self-interference in overlapping areas is analyzed by assuming a dependency on the received signal parameters. For this purpose, an experimental assessment is performed for creating a database that relates the received signal parameters to the resultant signal quality metrics. The laboratory setup emulates an SFN scenario with two interfering transmitters. The main received signal electric-field strength and the relative values of attenuation and delay corresponding to the remaining transmitter constitute the inputs in our system. The main received signal modulation error ratio (MER), the resultant signal MER, and the SFN gain are the output parameters. The proposed neural network models offer predictions of the main received signal MER, the resultant signal MER, and the SFN gain with a coefficient of determination (R2) of 0.99, 0.97, and 0.85, respectively. Several maps of SFN interference and signal quality are generated by using as input the multiple frequency network (MFN) data from the coverage software Radio Mobile. This procedure is also suitable by considering any MFN coverage software that returns signal strength and position data. The resultant self-interference can be predicted by considering a proper description of the MFN transmitter coverage, avoiding an expensive SFN test deployment.