基于VMD的瓦斯信号自适应压缩感知算法

(山东科技大学 计算机科学与工程学院,山东 青岛 266590)

瓦斯数据压缩; 压缩感知; 变分模态分解; 自适应观测矩阵; 信号稀疏化

Gas signal adaptive compressed sensing algorithmbased on VMD
WANG Tong-an,WANG Yuan-hong

(College of Computer Science and Engineering,Shandong University of Science and Technology,Qingdao 266590,China)

gas data compression; compressed sensing; variational mode decomposition; adaptive observation matrix; signal sparsity

DOI: 10.13800/j.cnki.xakjdxxb.2019.0226

备注

将压缩感知算法和变分模态分解相结合,应用于煤矿瓦斯数据的处理。考虑到现有的压缩感知算法在对瓦斯处理的过程中存在着重构精度低,重构过程复杂和需要较多的样本观测值等问题,因此提出一种基于VMD和自适应观测矩阵的压缩感知算法,有效解决了以较少的样本观测值数据实现信号高精度重构的问题,同时自适应地选择观测矩阵,避免了对稀疏信号的同类化投影选择。首先将瓦斯信号经过VMD进行分离,得到一系列瓦斯信号的本征模态函数分量,通过设定阈值保留有效信息,使得信号更加稀疏化; 其次通过自适应地观测矩阵对稀疏信号进行投影变换,从而降低观测矩阵和稀疏字典的不相关性。实验以煤矿瓦斯数据为研究对象,将瓦斯数据经过VMD分解进行稀疏化处理和使用构造的自适应观测矩阵进行投影选择,MATLAB仿真实验证明,文中的算法有更高的信噪比和更好的重构质量。

In this paper,the compressed sensing algorithm was combined with the variational mode decomposition to deal with the gas data in the process of coal mining.Considering that the existing compressed sensing algorithm had low reconstruction accuracy,complex reconstruction process and more sample observations in the process of gas processing,a compressed sensing algorithm based on VMD and adaptive observation matrix was proposed,which could effectively solve the problem of high reconstruction accuracy of signal with less sample observations data,and adaptively select the observation matrix to avoid the similar projection selection of sparse signals.Firstly,the gas signal was decomposed through VMD to gain a series of the Intrinsic Mode Function of the gas signal,and by setting the threshold,effective information was retained to make the signal more sparse; Secondly,the sparse signal was projected and transformed by the adaptive observation matrix,which reduced the correlation between the observation matrix and sparse dictionary.Taken the coal mine gas data as the research object in the experiment,the gas data was decomposed by VMD and the adaptive observation matrix was used for projection selection.MATLAB simulation shows that the algorithm has higher signal-to-noise ratio and better reconstruction quality.