Received signal strength computation for broadcast services using artificial neural network

Abstract

This paper investigates the influence of weather parameters on very high frequency (VHF) radio signals. Received signal strength (RSS) data were obtained from four broadcast stations in Niger State, transmitting at 91.2 MHz, 92.3 MHz, 100.5 MHz and 210.25 MHz while atmospheric parameters of temperature, pressure, relative humidity and wind speed data were obtained from the Tropospheric Data Acquisition Network (TRODAN) situated at the Federal University of Technology, Bosso Campus, Minna, Nigeria. The measurements were carried out for six months (January — July). An artificial neural network (ANN) model was designed to compute received signal strength using the measured atmospheric parameters. The training of the network was performed using Levenberg-Marquardt feed-forward backpropagation algorithm. The training process was performed by the evaluation of different effects of activation functions at the hidden and output layers, number of neurons in the hidden layer and data normalisation. The results obtained showed that the ANN model performed satisfactorily for the four broadcast stations as the computed signal strength values from the ANN model were reasonably close to the measured signal strength values with minimal errors. Also, the model performed well when tested on different data sets not used for the ANN training.