基于稀疏表示的系统辨识方法

基于信号的稀疏表示理论提出一种线性时不变系统辨识方法.该方法利用线性调频信号作为线性时不变系统激励输入信号,在利用传统方法进行系统辨识前利用稀疏分解算法对系统输出信号进行噪声处理.线性调频信号具有较好的时频聚集特性,线性时不变系统的输出也将具有很好的时频特征,利用基于Gabor字典的稀疏分解将能有效地提取输出信号中的有效分量,滤除其中的噪声成分,提高系统辨识的精度.仿真实验表明,本文提出的方法在低信噪比情况下,辨识效果好于传统方法.

System identification based on the sparse representation of signals

On the basis of sparse representation of signals, a novel method is proposed to identify the linear timeinvariant system in low signal-to-noise ratio environment. This method employs the chirp signals as the input to the identified system, and let the output be processed before identification by using the matching pursuit algorithm for noisereduction. Because of the time-frequency localization property of the input and output signals, a large amount of additive white noise can be reduced and the performance of system identification is thus improved. Simulation results show that the proposed method outperforms the conventional methods significantly in very low signal-to-noise ratio environment.