同频通信信号的分选是通信对抗侦察中经常遇到的问题，当各源信号互相独立时，独立分量分析(ICA)是进行信号分选的有效手段。介绍了ICA的基本理论，给出了ICA的数学模型和应用条件，详细分析了特征矩阵联合近似对角化(JADE)方法，提出将其应用到同频通信信号的分选中。在现有评价指标的基础上，针对通信信号提出将误码率差作为算法性能评价的一个新指标。最后将常用的数字通信信号作为源信号进行线性混叠，以此为观测信号进行了仿真实验，结果表明该方法不仅能分离不同调制样式的信号，还能分离相同调制样式的信号。

Sorting of overlapped communication signals is a common problem in communication countermeasure reconnaissance，and Independent Component Analysis (ICA) is an effective tool for signal sorting if the source signals are independent with each other. The basic theory of ICA is introduced, and the method of Joint Approximate Diagonalization of Eigenmatrix (JADE) is analyzed. The method was used in sorting of overlapped communication signals. It was proposed to use error rate as a new evaluation indicator. The common digital communication signals were taken as source signals for linear mixing, and the mixtures were used as observation signals for simulation. The result of simulation showed that the method can not only separate the signals with different modulation modes but also the signals with the same modulation mode.