基于字典学习的超宽带信号稀疏表示与降噪方法

Sparse representation and denoising method for UWB signal via dictionary learning

为了实现UWB(ultra-wide band)信号的高效稀疏表示,依据UWB信道的多径簇到达特性提出一种UWB信号稀疏表示方法。该方法利用S-V(saleh-valenzuela)统计模型以多径簇信号为原子设计冗余字典,由于多径簇模式下UWB信道的稀疏度小于单径模式,基于该冗余字典的UWB信号稀疏表示更简洁。给出构造冗余字典的误差加权递归最小二乘字典学习算法,在实现UWB信号稀疏表示的同时获得较高的信噪比增益,与同类算法相比总体性能较优。理论分析和仿真实验均证明了方法的有效性。

In order to solve the UWB signal effective sparse representation issue,this paper proposed the method by the multipath cluster arrival characteristic of UWB channel.According to the S-V statistic model,it constructed the redundancy dictionary by the proposed method using multipath cluster signal as atoms,and the UWB signal sparse representation could be effectively achieved as the UWB channel sparstiy defined by multipath cluster was less than that by single path.The error weighted recursive least square dictionary learning algorithm is then presented to construct the redundancy dictionary by which much improvement on UWB sparse representation SNR is obtained at the same time,and the improved performance over its counterparts was demonstrated using simulations.It verified the validity of the proposed method by both the theoretical analysis and the experimental results.