摘要
为解决稀疏度自适应匹配算法中使用内积法会造成部分原子信息丢失的问题,提出用Dice系数代替内积法进行原子相关度计算。使用Dice系数能够保留更多的原子信息,在重构信号时可以使结果更加精确。同时针对稀疏度自适应匹配追踪算法中使用固定步长导致稀疏度值过估计的问题,提出用信号残差控制步长大小的方法,将残差值分为2个区间,在不同的区间用2种不同的方式计算步长,使步长在残差处于第一区间时维持不变,让估计稀疏度值快速增长,达到第二个区间时使步长逐步减小,使估计稀疏度值缓慢贴近真实稀疏度,避免稀疏度过估计的情况,提升算法的重构精度。
In order to solve the problem of partial atomic information loss caused by the inner product method in the sparsity adaptive matching pursuit algorithm,the Dice coefficient is proposed to replace the inner product method to calculate the atomic correlation.Using Dice coefficient can retain more atomic information and make the result more accurate when reconstructing the signal.Besides,aiming at the problem of over estimation ofsparsity value caused by the fixed step size in sparsity adaptive matching pursuit algorithm,a method of controlling step size by signal residual is proposed.The residual value is divided into two intervals,and the step size is calculated in two different ways in different intervals.When the residual is in the first interval,the step size is fixed,so that the estimated sparsity value increases rapidly.When the residual reaches the second interval,the step size gradually decreases,so that the estimated sparsity value is slowly close to the real sparsity.In this way,the over estimation of sparsity value is avoided,and the reconstruction progress of the algorithm is improved.
作者
张世强
马可
张婷娟
黄雷
宋人杰
ZHANG Shiqiang;MA Ke;ZHANG Tingjuan;HUANG Lei;SONG Renjie(State Grid Heilongjiang Electric Power Co.,Ltd.Yichun Power Supply Company,Yichun 153000,China;School of Computer Engineering,Northeast Electric Power University,Jilin 132012,China)
出处
《黑龙江电力》
CAS
2020年第4期283-287,共5页
Heilongjiang Electric Power
基金
国家自然科学基金(项目编号:61271115)。
关键词
压缩感知
稀疏度自适应
Dice系数
重构算法
残差值
compressed sensing
adaptation of sparsity
Dice coefficient
reconstruction algorithm
residual value
