Practical Implementation of the FSK Demodulator based on Neural Network

Abstract:

Neural networks are used in every field of life, playing invaluable role in business. What is more, they are employed in the wireless communications area, including its physical layer in the form of demodulators. The reference works devote much space to theoretical studies of neural networks use to receive a modulated signal. Little space has been devoted to reference works containing practical implementations of NN-based receivers. This work proposes and studies a 4FSK modulation receiver based on a programmable radio USRP-2920. Demodulation based on the standard correlation demodulator was compared with the NN-based one. According to the studies, the neural network copes well with decoding modulated signal and is resistant to deformation introduced by the transceiver systems.