Spectrum Sensing in Cooperative Cognitive Radio Networks Utilizing Binarized Simplicial Convolutional Neural Network Performance in Dynamic Environments

ABSTRACT Cognitive radio (CR) is a new technology that aims to make better use of the radio spectrum. Its core component is spectrum sensing, which is difficult to accomplish precisely because many factors affect detection performance. Numerous spectrum sensing techniques, including the cyclic stationary detection algorithm, matching filter detection algorithm, energy detection algorithm, and others have been presented recently; these algorithms, though, depend on specific previous knowledge and are model‐driven. Inaccurate model assumptions or difficult‐to‐obtain prior knowledge will impair the algorithms' detection performance. A novel method of achieving spectrum sensing is made possible by the advancement of completely learning alongside neural network capabilities. In this paper, spectrum sensing in cooperative cognitive radio networks utilizing binarized simplicial convolutional neural network performance in dynamic environments (SS‐CCRN‐BSCNN‐DE) is proposed. Initially, the input signals are gathered from the GSM‐900 spectrum dataset. Then, the data are preprocessed utilizing reverse lognormal Kalman filtering (RLKF) to normalize the gathered signals. Then, the preprocessed data are fed into feature extraction using quadratic phase S‐transform (QPST) to extract the hidden spatial features. Then, the extracted features are fed into binarized simplicial convolutional neural network (BSCNN) for detecting the spectrum sensing in the cognitive radio system. The multiobjective thermal exchange optimization (MOTEO) is implemented to optimize the hyperparameters of BSCNN. The proposed method is implemented and its performance is evaluated under some metrics, like accuracy, signal‐to‐noise ratio (SNR), and root mean square error (RMSE), as well as computational time. The performance of the SS‐CCRN‐BSCNN‐DE approach attains 16.21%, 17.18%, and 24.44% higher accuracy; 21.25%, 17.25%, and 13.32% lower RMSE; 11.17%, 19.15%, and 23.14% lower SNR when compared with existing methods: cooperative spectrum sensing depending on convolutional neural network (CNN) in cognitive radio scheme (CSS‐CNN‐CRS), improved spectrum prediction method for cognitive radio networks utilizing artificial neural network (ISP‐CRN‐ANN), and recurrent neural network based spectrum sensing cognitive radio (RNN‐SS‐CR), respectively.