一种新的多径时延估计的约束反卷积方法

将lp范数规划方法和凸集交替投影(POCS)算法结合起来，利用一定的约束集，对接收信号进行多径分离，得到新的一种多径参数估计的约束反卷积方法，其中lp范数规划使得估计的多径结构具有一定的稀疏特性，然后利用POCS算法方便地引入各种约束，并保证估计结果收敛。仿真表明，这种方法比POCS算法和规划POCS算法具有更好的重构信号内部结构和参数估计的性能，能够分辨时延结构紧密多径分量，对噪声干扰表现也不敏感。新方法无需测量信道的实际物理参数便可清晰地分辨能量较大的多径成份，利用这种方法成功地进行了海试数据的分析处理，得到的多径结构和声线传播模型一致。

A New Constrained Deconvolution Method and Its Application in Multipath Time-Delay Estimation

Refs.4 and 5 presented constrained deconvolution methods that are, in our opinion, not always effective for multipath time-delay estimation. In this paper, we propose a new constrained deconvolution method that can always do so. Our new method integrates l_p-norm regularization with POCS (Projection Onto Convex Sets) algorithm. The introduction of l_p-norm regularization into our new method does two things: (1) to remove the difficulty caused by the sparsity of multipath structure; (2) to resolve closely-spaced multipath components. Simulation results show preliminarily that our new method shows improvements in two respects: (1) the multipath time-delay structure estimated by new method is always sparse; (2) estimated parameters, both in structure and in numerical value, approach sufficiently close to preassigned values and relatively more accurate results are obtained. We also collected data from sea tests and analyzed them both with our new method and with the method of acoustic-ray-propagation model. The comparison of the results obtained with the two methods of analysis shows that our new method is better in that multipath components with relatively more energy can be clearly distinguished without requiring real-time measurement of physical parameters.