Gravitational Search Algorithm Based Probabilistic Neural Network for Spectrum Sensing in Cognitive Radio Networks

Abstract

Spectrum sensing in cognitive radio networks is a promising technology and has been done with the help of different prediction models and probability theory-based algorithms available in the literature.  There are some works which employ the neural network for predicting the spectrum availability, but has drawbacks related to training algorithm employed which leads to inaccurate spectrum sensing. For overcoming these limitations in spectrum sensing in a cognitive radio environment, AI-based techniques and optimization procedures are being increasingly employed. In this paper, the Gravitational Search algorithm based Probabilistic Neural Network (GS-PNN) is utilized for range detecting in cognitive radio systems. In the procedure, the client information is transmitted through the inert or abandoned diverts as of now in the framework which is found out using GS-PNN. The proposed technique is evaluated using the parameter graphs of, SU_imp and throughput. The comparative analysis is carried out by comparing proposed technique results to HMM, LM based NN, GS-LM and random technique. From the analyses, we can infer that the proposed technique achieved results by having higher evaluation metric values. The highest, SU, SU_imp and throughput values achieved by our technique are about 0.585, 0.52 and 1.36 respectively.