Exploring Advanced Approaches: A Comprehensive Analysis of Machine Learning and Deep Neural Networks in Spectrum Sensing Applications

The increasing demand for wireless communication resources has intensified the need for effective spectrum sensing techniques to alleviate the scarcity of available spectral bands. This survey paper presents an indepth exploration and comparative analysis of various spectrum sensing methodologies, focusing on Machine Learning (ML) and Deep Neural Network (DNN) approaches. The study delineates the landscape of traditional spectrum sensing techniques, encompassing energy detection, matched filtering, cyclo-stationary feature detection, and cooperative spectrum sensing. This paper explains sensing using machine learning-based such as Support Vector Machines (SVM), Random Forest, Neural Networks, and Hidden Markov Models (HMM) are systematically elucidated. Moreover, the paper meticulously scrutinizes the advancements in deep learning-based spectrum sensing, encapsulating the utilization of Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Deep Reinforcement Learning (DRL) methodologies. The discussion encompasses their applications, advantages, and limitations in the context of spectrum sensing.