Abstract:

This work proposes a QAM demodulator based on neural networks. The demodulator operation was analysed for two neural networks, i.e. a simple Multilayer Percepron (MLP) and a Convolutional Neural Network (CNN). The transmitted signal is subject to the activity of the channel with Rayleigh fading. The obtained results were compared with the ones obtained using a standard QAM demodulator. The simulation results indicated that the neural network for the preset propagation model offers better results calculated by means of the bit error rate (BER).