Design of Adaptive Beam Former Using 2 Dimensional Lattice Filters and Its Performance Verfication

(1)提出了一种新的自适应波束形成方法，该方法对信号模型误差具有较强的鲁棒性，并减少了计算量。为了减小模型误差的影响，本文基于时域特征向量约束校准广义旁瓣对消器的方法，推导了阻塞矩阵和传递矩阵。(2)提出了一种用于说话人波达方向估计的麦克风阵列系统。它利用谐波成分，如元音。它在方向性噪声和传感器噪声下是有效的。(3)针对DS/CDMA系统中的频率选择性衰落和共信道干扰抑制问题，提出了一种基于GMA自适应阵列和抽头延迟线的接收机。本文研究了三波约束恒模算法的收敛性。基于分析结果，我们提出了一种新的标准来选择成功捕获所需信号的模因子。(4)提出了一种基于线阵Hopfield网络的波达方向估计方法。该方法对传感器误差具有鲁棒性。此外，该方法对分布式信源的DOA估计也有一定的应用价值。这些方法即使在低信噪比的情况下也能给出正确的方向。(5)提出了一种基于径向基函数（RBF）网络的数字信道均衡器，该网络与最优贝叶斯符号检测具有相同的结构。提出了一种聚类算法来寻找正确的径向基函数中心。仿真结果表明，该算法比传统算法具有更好的收敛性和更少的计算量。(6)提出了一种利用输入相关函数自适应搜索梯度的新算法。该算法用于电话会议系统中的回声消除。

(1) A new method of adaptive beam forming which is robust against signal model errors and reduces computational load is established. In order to reduce the affection of the model errors, we deduce both blocking matrix and passing matrix based on the method of eigenvector constrained calibrated Generalized Sidelobe canceller in the time domain.(2) The microphone array system to estimate direction-of-arrival of speaker is proposed. It utilizes harmonic component such as vowel sound. It is effective under a directional noise and sensor noise.(3) A receiver using GMA adaptive array combined with Tapped Delay Line is proposed for frequency selective fading and co-channel interference suppression in DS/CDMA systems. We investigate the convergence behavior of the constrained CMA with three waves model. Based on analytical results, we propose a novel criterion to select a modulus factor of capturing desired signal successfully.(4) An estimation method of direction-of-arrival using Hopfield network with linear array is proposed. The proposed method is robust for sensor error. Additionally, the method is useful to estimate DOA of distributed source. These methods can give proper direction even the low SNR condition.(5) It is proposed that a digital channel equalizer using radial basis function (RBF) network which has identical structure to the optimal Bayesian symbol-detection. A clustering algorithm has been proposed to find the correct center of RBF. The simulation results show its better convergence and less calculation than the conventional algorithm.(6) A new algorithm for tap adaptation to search gradient by input correlation functions is proposed. This algorithm is used for echo cancellation in teleconference systems.