Integrated Blind Signal Separation and Neural Network Based Energy Detector Architecture

Abstract

Energy detector is the simplest spectrum sensing technique in cognitive radio in terms of computation complexity as it is developed assuming only channel noises for the detection of primary user signal. However in realistic sense the typical communication system comprises of various other sources of noise (like thermal noise) that significantly affect the sensing performance of the detector. The conventional energy detector is also not a coherent detector as the attributes of the primary user signal is unknown. Considering these limitations a system is proposed that can have the knowledge of the primary user signal attributes under realistic situation at the secondary user (SU) receiver thereby enhancing the accuracy of the sensing significantly. In this work a new coherent energy detector system architecture is proposed that first separates the signal components from a noisy signals (non Gaussian noise) at SU receiver, then from the separated signals it estimates the signal attributes (power of the signal, noise variance, signal-to-noise ratio) required for conventional energy detector. The work is developed as a hybrid system model utilizing blind source separation algorithm and neural network (NN) known hereafter as ‘integrated blind signal separation (BSS)-NN based energy detector’.