Space-Time Processing for Digital Comm: Nonparametric Channel ID & Interference Cancellation & Multichannel Equalizer Design Based on Linear Matrix Inequalities

在城市蜂窝环境中，由于建筑物和其他结构的反射，数字通信中使用的数据脉冲通常通过从发送器到接收器的多条路径传播。这种多径传播导致每个脉冲被在其之前和之后传输的脉冲所破坏；这种“符号间干扰”(ISI)的有害影响需要以“均衡器”的形式进行信号处理。更复杂的是，预计蜂窝用户数量的爆炸式增长将几乎肯定需要相邻小区中的两个用户占用相同的频带。这会导致“同频道干扰”，表现为用户之间的串扰。天线阵列的使用能够区分来自不同方向的信号，提供了抑制同频道干扰的手段。本研究致力于发展时空信号处理方案，以在城市蜂窝环境中实现移动窄带数字通信的均衡和干扰消除。将均衡器设计问题转化为基于线性矩阵不等式的凸优化问题，并进行了数值求解。均衡是通过在每个天线上采样间隔抽头来实现的，包括大致等于多径时延扩展的持续时间。与使用符号间隔抽头的传统均衡方案相比，这有助于更好地跟踪时变多径信道。通过形成具有朝向期望用户的多径的最大增益方向和朝向强同频干扰的零点的波束来实现干扰消除。这是通过通过一种新颖的扩展相关器处理短训练信号来实现的，该扩展相关器在时间上高度定位期望用户的信号的贡献，从而允许估计干扰的空间相关矩阵。对于时分多址(TDMA)系统，如基于IS-136 TDMA标准的蜂窝系统，在多径时延扩展小于脉冲符号波形的零到零主瓣的情况下，本研究中开发的均衡和干扰消除方案在收敛速度、计算复杂性、对模型失配的稳健性和符号错误性能之间的权衡方面具有最大的前景。

In an urban cellular environment, the data pulses employed in digital communications typically travel through multiple paths from transmitter to receiver due to reflections from buildings and other structures. This multipath propagation causes each pulse to be corrupted by pulses transmitted before and after it; the deleterious effects of this "intersymbol interference" (ISI) necessitate signal processing in the form of an "equalizer". As a further complications, the projected explosion in the number of cellular users will almost certainly require two users in the same of adjacent cells to occupy the same frequency band. This leads to "co-channel interference" that manifests itself as cross-talk amongst users. The use of antenna arrays, which enables discrimination amongst signals arriving from different directions, offers a means of suppressing co-channel interference. This research focuses on the development of spatiotemporal signal processing schemes to effect both equalization and interference cancellation for mobile narrowband digital communications in an urban cellular environment. The equalizer design problem is reformulated as a convex optimization problem based on linear matrix inequalities and solved numerically. Equalization is effected with sample-spaced taps at each antenna, encompassing a time duration roughly equal to the multipath time delay spread. This facilitates better tracking of time-varying multipath channels as compared to conventional equalization schemes employing symbol-spaced taps. Interference cancellation is effected by forming beams with directions of maximum gain steered towards the desired user's multipath, and nulls steered towards strong co-channel interferers. This is achieved by processing a short training signal via a novel extended correlator that highly localizes in time the contribution of the desired user's signal, thereby allowing one to estimate the spatial correlation matrix of the interferers. The equalization an d interference cancellation schemes developed in this research hold the greatest promise in terms of the trade-off amongst convergence rate, computational complexity, robustness to model mismatch, and symbol error performance for Time Division Multiple Access (TDMA) systems where the multipath time delay spread is less than the null-to-null mainlobe of the pulse symbol waveform, as is the case with cellular systems based on the IS-136 TDMA standard.