| 摘要: |
| 信号分离是雷达电子对抗的重要环节。考虑到雷达信号在时频域具有稀疏性的特点，在独立分量分析的基础上，提出了一种基于时频域稀疏性的线性调频雷达信号盲源分离方法。首先对混合信号进行短时傅里叶变换，在每个频点利用自然梯度算法分离信号，由分离信号幅度的比值作为对源信号后验概率的估计；然后根据相邻频点后验概率序列的相关性进行排序，确保各个频点的分离信号属于同一个源信号；最后设计时频掩码分离信号。进行了线性调频雷达信号卷积混合的盲分离实验，所提方法分离结果明显优于传统独立分量分析方法的分离结果，验证了该方法的有效性。 |
| 关键词: 线性调频信号 盲源分离 卷积混合 稀疏性 后验概率 时频掩码 |
| DOI： |
|
| 基金项目: |
|
| Blind source separation of LFM signals using time frequency domain sparseness |
| CHEN Anjun,YANG Qingshan,WANG Jie,YUAN Kang,YIN Fuliang |
| () |
| Abstract: |
| Signal separation is vital for radar electronic countermeasures.According to the sparseness of radar signals in the time-frequency(T-F) domain,a T-F domain sparseness based linear frequency modulation(LFM) radar signal blind source separation(BSS) method is proposed based on the independent component analysis(ICA) method.Firstly,with the short-time Fourier transform(STFT) of the mixed signals,natural gradient algorithm is used to separate signals in each frequency bin,and the ratio of magnitudes of the separated signals is taken as the estimation of posterior probability.Secondly,the permutation ambiguities are aligned according to the correlations of posterior probability sequences in adjacent frequencies to ensure that the separated signal in each bin belongs to the same source.Finally,T-F mask is designed to separate signals.BSS experiments are conducted for convolutively mixed LFM radar signals and the proposed method’s separation results are obviously better than those of traditional ICA method,which shows the effectiveness of the proposed method. |
| Key words: linear frequency modulation signal blind source separation convolutive mixtures sparseness posterior probability time-frequency mask |
