中文摘要：

针对水声信号处理中常见的多途、相干信号的多目标分辨和方位估计问题，依据遗传算法突出的求解线性及非线性方程（组）数值解的优势，研究了相干信号的盲分离和方位估计方法．用“L”型阵列接收多个相干目标信号，在建立阵列流形和盲分离数学模型的基础上，利用复数遗传算法求解非线性方程组进行各相干目标源的分离，并利用阵型优势解决线列阵在方位估计上的左右舷模糊问题，对各相干目标源进行方位估计．MATLAB仿真试验表明：该分离算法能够快速收敛于全局最优解，成功实现相干信源的盲分离，最终实现多个相干目标的分辨及其方位估计．

英文摘要：

While aiming at multi-object identification and DOA estimation problems of muhi-channel coherent signals in acoustic signal processing by utilizing the prominent advantages of genetic algorithms for solving the numerical solutions of linear and nonlinear equation groups, a blind separation and DOA estimation method based on coherent signals was studied. An ＂ L＂-type array was used to receive many coherent source signals, and coherent sources were separated by using complex genetic algorithms for solving linear equation groups on the basis of establishing the array manifold and mathematical model of blind source separation. In addition, the left and right sides of the fuzzy problem in DOA estimation of linear arrays were solved by using array formation advantages in order to estimate direction of coherent sources. MATLAB simulation results show that the separation algorithm can quickly converge to the global optimal solution, successfully separate coherent sources, and ultimately realize identification and DOA estimation of coherent objectives.