摘要
在信号去噪问题中,利用K-means Singular Value Decomposition(K-SVD)等经典字典学习算法,对信号进行稀疏分解与信号重构,不能有效的消除噪声影响。引入了非线性最小二乘和粒子群优化的方法对经典的字典学习去噪方法进行了改进。利用K-SVD算法进行字典训练;利用非线性最小二乘的方法对字典中的每一个原子进行拟合,得到修正后的字典;利用粒子群优化的方法求解信号的稀疏表示,最终得到重构信号。通过实验证明,该方法去除噪声的效果相较于K-SVD和RLS-DLA(递归最小二乘字典学习算法)有明显提高。
In signal denoising problems, using K-SVD and other classic dictionary learning algorithm can not effectively eliminate the noise impact. The method made some amendments for classical dictionary learning by applying nonlinear least squares curve fitting and particle swarm optimization. K-SVD algorithm was used to train the dictionary. Nonlinear least-squares approach was used to fit every atom in the dictionary. Particle swarm optimization method was used to solve the sparse representation of the signal. The reconstructed signal was obtained. The experimental results show that, the denoising effects of the proposed method apparently has increased compared with K-SVD and RLS-DLA.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2015年第12期2935-2941,共7页
Journal of System Simulation
基金
国家自然科学基金(61372136)
关键词
字典学习
去噪
粒子群优化
信号重构
曲线拟合
dictionary learning
denoising
particle swarm optimization
signal reconstruction
curve fitting
