提出了一种基于Gabor变换的盲信号分离方法。与传统的盲信号分离方法相比，该方法考虑到了不同类型信号的时频分布特点，通过估计混合矩阵从而能够较为准确地对源信号进行分离，而且突破了传统盲信号分离方法中要求源信号相互独立以及源信号最多只能有一个高斯信号的限制。仿真结果验证了该方法的有效性，为盲信号分离技术提供一种新的研究方向。

A new method of Blind Source Separation (BSS) based on Gabor transform is proposed. Compared with the traditional BSS methods, this new method could obtain the estimate of the mixing matrix so as to separate the mixing signals based on the characteristics of time-frequency distribution of the source signals. Even if the source signals are correlative, or there are more than one Gaussian signals in the sources, one could get better separation performance by this new method. The simulation results show the feasibility and validity of the method, which may provide a new research direction for blind signal separation.