文献类型：期刊文献

中文题名：基于压缩感知的高分辨频率估计

英文题名：High Resolution Frequency Estimation with Compressed Sensing

作者：刘兆霆[1,2];何劲[1];刘中[1]

机构：[1]南京理工大学电子工程系;[2]绍兴文理学院电子工程系

年份：2009

期号：8

起止页码：1252

中文期刊名：信号处理

外文期刊名：Signal Processing

收录：CSTPCD、、CSCD2011_2012、北大核心2008、北大核心、CSCD

语种：中文

中文关键词：压缩感知;分辨率;二次锥规化;稀疏分解

外文关键词：Compressed Sensing ; Resolution ; SOCP ; Sparse Representation

中文摘要：压缩感知是信号离散表示的新理论。本文将该表示理论用于正弦信号的频率估计,提出一种新的高分辨率的频率估计方法。该方法根据信号的稀疏表示,利用一个随机的压缩矩阵先对信号进行压缩,再在压缩域中通过对l_1模优化重构该稀疏信号,获得信号的频率估计。模拟分析了新方法性能,并与直接l_1模优化算法、Pisarenko、MUSIC等算法进行了比较。结果表明本文方法分辨性能明显优于Pisarenko和MUSIC等算法;具有直接l_1模优化算法相当的性能,但计算量大大降低。

外文摘要：Compressed sensing is a new theory of discrete signal representation. This paper applies the theory to the frequency estimation of sinusoidal waves contaminated in noise and develops a new high resolution method. The proposed method firstly compresses the signal with a random compressed matrix and then estimates the signal frequencies in the compressed domain by l1 -norm optimization. In comparison with related methods, the new method outperforms greatly Pisarenko and MUSIC methods and has the performance comparable to the direct l1-norm optimization method. However, the new method has the computational advantages because the fre quency estimation is conducted in the compressed domain.