Neural network prediction of radio propagation

Preliminary work for predicting signal distributions in a local area, using a feature-based neural network is presented. The neural network is trained by radio signal measurements at known positions. After appropriately setting the parameters of nodes of the neural network, a corresponding virtual propagation environment is built, which reasonably represents the actual environment. Radio signal strength distribution is predicted by the virtual environment. A new method for radio signal measurement is introduced which mitigates the effect of small scale fading when determining the fingerprint of a position