Robust machine-learning based algorithm for detection of signal under noise floor

Abstract

Spectrum sensing plays an important role in cognitive radio. In wireless communication systems, due to severe transmission environment of interference, the received signals may be very weak as compared to the background noise. In this project, first, the existing schemes of detection of signals below the noise floor are studied. Following that, a machine-learning based algorithm using one-dimensional convolution neural network is developed and applied to detect the presence of signals below the noise floor. By testing on various cases and comparing with existing methods, it shows better performance and higher accuracy. It also brings out potential study subjects concerning real life application and signal enhancement.