非高斯脉冲干扰下MIMO系统频谱智能感知方法

An Intelligent Spectrum Sensing Method for MIMO System with Non-Gaussian Impulsive Interference

智能频谱感知是未来电磁频谱管控的关键支撑技术.针对现有多输入多输出系统(multiple-input mulitiple-output,MIMO)频谱感知方法提取特征能力有限,且无法适应非高斯脉冲干扰等问题,提出一种基于注意力机制的协作频谱智能感知方法.通过引入非线性变换器抑制非高斯脉冲干扰,并利用广义协方差矩阵表征MIMO信号统计特性,将其作为后续网络的输入.基于Transformer模型构建特征提取网络,通过其自注意力机制挖掘不同的深层特征.采用随机样本对网络进行训练,并利用网络输出特征向量设计二元假设检验方法实现MIMO系统的频谱智能感知.仿真结果表明,所提方法在高斯噪声和非高斯脉冲干扰下可以有效实现MIMO系统频谱感知,且在低信干比条件下具有较高的感知性能.

Intelligent spectrum sensing is a key supporting technology for the future electromagnetic spectrum management and control.Aiming at the problems that existing spectrum sensing methods for MIMO systems have a limited ability to extract features and cannot adapt to non-Gaussian impulsive interference,etc.an intelligent spectrum sensing method for coordinated spectrum based on attention mechanism is proposed.A nonlinear transformer is introduced to suppress nonGaussian impulsive interference,and a generalized covariance matrix is used to characterize the statistical properties of the MIMO signals and to act as the input to the subsequent network.A Transformer model-based feature extraction network is then constructed to mine different deep features through its self-attention mechanism.The network is trained with random samples and a binary hypothesis testing method is designed to achieve spectrum sensing using the network output feature vectors.The simulation results show that the proposed method can effectively achieve MIMO spectrum sensing under Gaussian noise and non-Gaussian impulsive interference,and has a higher sensing performance under low signal-tointerference ratio.