基于贝叶斯稀疏重构的非均匀样本谱估计

在航空电子对抗领域,往往需要利用非均匀样本来估计信号的频谱。针对非均匀样本谱估计问题,提出了贝叶斯稀疏重构谱估计算法(BSRSE)。该算法首先将非均匀采样的谱估计表示为稀疏信号重构问题。然后利用拉普拉斯分布表示稀疏性,建立贝叶斯模型。最后通过构造加速的不动点迭代方法估计参数,从而估计信号频谱。与现有谱估计方法比较,该算法具有较高的频率分辨力、较强的噪声适应能力,且需要较少的样本数。数值仿真验证了该算法的有效性。

Spectrum estimation of nonuniformly sampled data based on Bayesian sparse recovery

In the field of aviation electronic warfare,the signal spectrum often needs to be estimated using nonuniform samples.A novel approach,called Bayesian sparse recovery spectral estimation（BSRSE）,is proposed to solve the problem of spectrum estimation of nonuniformly sampled data.Firstly,spectrum estimation is cast as sparse recovery problem.Then,Bayesian model is constructed using the Laplace priors to enforce the sparity.Finally,a fast fixed point method is used to estimate parameters to obtain the spectrum.The proposed algorithm has some advantages over most existing methods：it improves the frequency resolution,it is more robust to noise,and it needs less nonuniformly sampled data.The effectiveness of this algorithm is demonstrated by simulations.