中文摘要：

针对混叠跳频信号盲分离中特征矩阵对应的自项时频点选取不准和联合对角化正交约束性问题,提出了一种基于组合时频分析的跳频信号盲分离算法。该算法先提出Gabor＆SPWVD组合时频分析,利用其时频图清晰的优点准确提取具有特征矩阵结构的自项时频点,然后构造自项时频点的空间时频矩阵,并利用最小二乘原则对其进行非正交对角化,最后实现分离矩阵的估计以及多个混叠跳频信号的盲分离。仿真结果表明,该算法不仅能有效实现不相关跳频信号的盲分离,还可以实现相关跳频信号的分离,与TFBSS算法在不同信噪比条件下的分离性能相比具有更好的抗噪能力。

英文摘要：

Aiming at the problems of the auto-source time-frequency points selection not accurate and the orthogonal binding of joint diagonalization in blind separation of frequency-hopping signals, a new blind separation algorithm based on the analysis of combination time-frequency was proposed. The algorithm effectively obtained the auto-source （time-frequency） TF points with the Gabor＆SPWVD combination time-frequency analysis and compute a sequence of matrices of time-frequency distributions （TFDs） , and then, the separate matrix was esti-mated through non-orthogonal joint diagonalization realizing blind source separation of mixed frequency-hopping signals. Finally, the simulations illustrated that the proposed algorithm was effective in the blind separation of frequency-hopping signals whether relevant or not, and had good anti-noise performance compared with TFBSS algorithms of frequency-hopping signals.