基于信号循环平稳特征的神经网络频谱感知算法

Spectrum sensing algorithm based on neural networks with cyclic autocorrelation characteristics

高效、准确的频谱感知是实现认知无线网络的重要前提。文中针对低信噪比环境下的认知无线网络频谱感知问题,研究基于主用户调制信号循环平稳特征的神经网络频谱感知算法。在该算法中,将信号循环平稳特征的感知网络与神经网络的特征输入层有机结合,构造一个全连接的人工神经网络,并在神经网络的输出层连接一个Soft-max网络,对频谱进行判决。该算法不需要预设频谱判决门限,有效降低了噪声波动对频谱感知性能的影响。仿真结果表明,在低信噪比和噪声波动等复杂信道环境下,该算法具有良好的频谱感知性能。

The efficient and accurate spectrum sensing is important for implementing cognitive radio networks.The spectrum sensing of cognitive radio networks in the low signal to noise ratio(SNR)is studied and a novel spectrum sensing algorithm is proposed based on the cyclostationary characteristics of the modulated signal of primary user and neural networks.In the algorithm,a fully connected artificial neural network is constructed by combining the sensing network of the cyclostationary characteristics with the input layer of the neural network.The output layer of the neural network is connected to a Soft-max network for the spectrum decision.The algorithm does not need to preset the threshold of the spectrum decision,thus reducing the effect of noise fluctuation on the spectrum sensing performance.Simulation results show that the algorithm has good performance in the envirenment with the low SNR and the noise fluctuation.