复杂电磁环境下,雷达、通信等信号在时域、频域、时频域存在复杂混叠,且各类型信号带宽差异大、调制样式多样,常规信号分离方法难以应用。文中提出了一种基于信号重构的非参数化混叠信号分离方法。基于瞬时幅度的傅里叶分解,建立非参数化稀疏信号模型,将混叠信号分离转化为瞬时幅度和瞬时频率的联合估计问题,基于交替迭代思路,分别进行估计和更新。对于瞬时幅度项,基于广义近似消息传递方法进行估计;对于瞬时频率项,利用瞬时频率变化平缓性特征,建立瞬时频率更新模型,实现对瞬时频率的更新。进一步,基于瞬时幅度和瞬时频率估计结果进行各信号重构,实现混叠信号分离。仿真结果表明:文中所提方法能有效分离时频域复杂混叠信号。

In complex electromagnetic environment, radar and communication signals are mixed in the time, frequency and timefrequency domains. The signal bandwidth varies widely and there are diverse modulation modes. Conventional source separation methods are difficult to be effectively applied. In this paper, a reconstruction-based non-parametric source separation method of aliasing signals is proposed. A non-parametric sparse model of received signal is established based on Fourier decomposition of the instantaneous amplitude (IA). The source separation of the aliasing signal is transformed into joint estimation of the IA and instantaneous frequency (IF). Based on the idea of alternate iteration, the estimation and update of IA and IF are carried out respectively. The IA of the signal is estimated based on the generalized approximate message passing. For the IF term, an update model of the IF is established based on the gentle variation to realize the updating of IF. Further, based on the estimation results of the IA and IF, each sub-signal is reconstructed to achieve aliasing source separation. Simulation results show that the proposed method can effectively separate the complex aliasing signals in the time-frequency domain.

TN957.51

国家自然科学基金资助项目;江苏省自然科学基金资助项目