两类混合特征信号的超完备稀疏表示方法

Overcomplete and Sparse Representation of Two Kinds of Signals with Mixed Features

本文提出了基于跳跃字典的超完备稀疏表示方法和基于自适应分割定义域的超完备稀疏表示方法,分别用于重建带有周期和方波特征的信号和带有周期和冲击特征的信号.实例表明,这两种方法对相应的信号在逼近误差和稀疏性上达到了比直接采用基追踪或小波逼近更好的效果.

This paper addresses new methods to recreate signals with periodic and rectangular features, employing Jump Dictionary-based overcomplete sparse representation method, and signals with periodic and impulse features, employing self-adaptive segmentation-based overcomplete sparse representation method. As revealed in our trials, in terms of approximation error and sparsity, both of the two methods mentioned above get better prediction results than basis pursuit algorithm and wavelet approximation.