蜂群优化神经网络的频谱感知

A Neural Network Spectrum Sensing Algorithm Using Bee Colony Optimization

由于神经网络训练收敛速度慢、易陷入局部最优解,而最优解对神经网络的频谱感知算法性能影响大,因此为提高神经网络的频谱感知算法性能,采用蜂群算法交叉训练神经网络,加快训练收敛速度,降低均方误差。采用信号的能量、循环功率谱作为特征参数,提出了蜂群优化神经网络的频谱感知算法。仿真结果表明,在给定迭代次数下,相比能量法、循环平稳特征法、无蜂群算法交叉训练神经网络或RBF神经网络的频谱感知算法,本文算法具有更好的感知性能。

Neural network is easy to converge on local optimal solution, and its training convergence speed is slow. In ad- dition, the optimal solution has a deep influence on performance of a spectrum sensing algorithm of neural network. There- fore, neural network is cross-trained by bee colony algorithm to accelerate the training convergence speed and to reduce the mean square error so as to improve the performance of a neural network spectrum sensing algorithm. Using signal energy and spectral correlation as feature parameters, a neural network spectrum sensing algorithm using bee colony optimization is proposed. Simulation results show that with a certain number of iterations, the proposed algorithm has better sensing per- formance compared with the spectrum sensing algorithm based on energy detection, the spectrum sensing algorithm based on cyclostationary, the neural network spectrum sensing algorithm without bee colony cross training and spectrum sensing algo- rithm based on RBF neural network.