摘要
复杂噪声背景中经典的谐波恢复方法存在分辨率不高或噪声分布的假设很强的局限性。针对上述问题,提出引入稀疏分解的思想,采用了一种新的正交匹配追踪的谐波恢复方法。应用余弦原子模型,并依据循环统计量初估计结果生成过完备原子库。根据乘性噪声均值是否为零分别提出了相应的算法。在较低的噪声假设条件下,仅根据噪声之间的相关性得到了分辨率较高的方法。仿真结果说明新方法的有效性。
Classical harmonic retrieval algorithms in complex noises suffer the problems of low frequency resolution and unrealistic assumption. In order to deal with this issue, this paper proposed a novel matching pursuit based method. The new method applied cosine atom model and the dictionary was constructed by using of the results obtained by cycle -based algorithm. The correspondiog algorithms were designed according to whether the mean value of the muhiplicative noise is zero. The assumption in the new method is mild and practical compared to those in classic algorithms. Some numerical results were presented to confirm the validity the new method.
出处
《计算机仿真》
CSCD
北大核心
2013年第2期178-182,256,共6页
Computer Simulation
基金
国家自然科学基金(61102103
61071188)
湖北省自然科学基金(2010CDB04205)
中央高校基本科研业务费专项资金资助(CUG090112
CUG110407
CCNU10A01013)
关键词
谐波恢复
高分辨率
复杂噪声
正交匹配追踪
Harmonic retrieval
High resolution
Complex noises
Orthogonal matching pursuit（OMP）
