Dandelion Exponentially Weighted Moving Average enabled Siamese Convolutional Neural Network for Cooperative Spectrum Sensing in Cognitive Radio Network

Sensing of spectrum is significant for cognitive radio systems for authorized users to avoid interruption. It also finds spectrum by improving usage of the system. However, in real-world applications, detection performance of system can be affected by some issues like shadowing, receiver uncertainty and multipath fading. To address these issues, Cooperative Spectrum Sensing (CSS) is utilized and it also uses method of spatial diversity to improve the detection process in wireless networks. CSS offers benefits such as reduced sensitivity needs and better detection mechanisms. It also faces cost expensive issues, such as delays, operations, increased sensing time and usage of energy, and also potential performance reduction. Here, Dandelion Exponentially Weighted Moving Average enabled Siamese Convolutional Neural Network (DEWMA_SCNN) is devised for CSS in Cognitive Radio Network (CRN). Initially, the system model is considered. Then, test statistics for CSS is performed. Then, combine the decisions of components of signal, such as Test statistics, Eigen statistics, Signal energy, Wavelet transform, Cyclostationary feature, Compressed sensing feature and Matched filter. Moreover, Siamese Convolutional Neural Network (SCNN) is utilized for generating final decision, which is fine-tuned using Dandelion Exponentially Weighted Moving Average (DEWMA). Here, DEWMA is the combination of Dandelion Optimizer (DO) and Exponentially Weighted Moving Average (EWMA). Furthermore, the performance of proposed approach is assessed with metrics, such as probability of false alarm, sensing time, and probability of detection has achieved the values 0.458, 0.666, and 279.764 sec, respectively.