自适应宽带稳健波束形成及神经网络算法

针对被动跟踪系统中阵处理的要求，建立了适应宽带稳健波束形成设计方法，用神经网络算法实现了自适应宽带波束形成，从理论上解决了自适应阵观察方向上的稳健设计问题，抛弃了自适应阵的时延补偿器，将正则网络应用于自适应宽带波束形成，解决了阵处理中求二次约束解的非线性规划问题，并能在毫微秒时间内收敛，不仅保持了二次约束自适应波束形成的干扰抵消能力，而且克服了软约束带来的阵指向性两端上的现象，所建立的阵处理器能抵消干扰达40dB以上，工作带宽达到2个倍频程。

Robust adaptive broadband beamforming and neural network

A robust adaptive broadband beamforming method for the array processing in underwater passive tracking system is introduced. The neural network is applied to the broadband beamforming and the problem of the robust design on searching directions in adaptive array processing is solved theoretically. The method is robust and avoids delay compensating for each element. This paper applies the canonical neural network to the adaptive beamforming, and solves the problem of the nonlinear programming with quadratic constraint in the broadband array processing. This broadband beamformer keeps the performance of the adaptive beamformer with quadratic constraint to cancel interference, and can be convergent within nanoseconds. It avoids the raising in two sides of the array directions which appeared with soft constraint, and can cancel the interference up to 40 dB or over in two octave bands with the canonical nonlinear programming circuits.