Narrowband Spectrum Sensing: Fuzzy Logic Versus Deep Learning Systems

Abstract

The motivation for this work was to investigate the advantages and disadvantages of two promising techniques for narrowband spectrum sensing: fuzzy logic and deep learning which can be useful for future users. To this end, we present three fuzzy logic systems and four deep learning-based systems for narrowband spectrum sensing. The fuzzy logic systems include triangular and Gaussian membership functions, multiple implications, and aggregation methods. The deep learning systems are based on three basic architectures, including convolutional neural networks (CNN), long short-term memory (LSTM), and fully connected (FC) layers. Simulation results show that deep learning techniques provide a higher probability of detection in a wider SNR range than fuzzy logic techniques. However, fuzzy logic utilizes simpler hardware-friendly detectors, than deep learning.