Spectrum Sensing Algorithm Based on Twin Support Vector Machine

Abstract

Spectrum sensing is the key of implementing cognitive radio technology. As a kind of machine learning method based on statistical learning theory, support vector machine has the advantages of global optimization, nonlinearity and good generalization ability. The use of support vector machines in spectrum sensing can solve the problem that parameters in spectrum sensing are difficult to determine by learning historical data. Aiming at the problem that the training time of the spectrum perception algorithm based on support vector machine is too long, this paper proposes a spectrum sensing algorithm based on twin support vector machine based on fuzzy mathematics and twin support vector machine. The algorithm extracts the leading eigenvector as the input features that can reflect the signal correlation and calculate the membership degree according to the proportion of the maximum eigenvalue. The twin support vector machine is used to solve the classification hyperplane. The algorithm complexity is only 1/4 of the SVM algorithm, which can greatly reduce the training time. The simulation results show that under equal prior condition, when the number of users and the number of samples are the same, the spectrum sensing algorithm based on TWSVM can obtain a lower minimum error probability than the SVM-based spectrum sensing algorithm and energy detection. As the number of users and the number of sampling points increase, the minimum error probability of the TWSVM-based spectrum sensing algorithm decreases.