基于稀疏表示的数据无失真压缩模型构建  

Construction of data lossless compression model based on sparse representation

作　　者：孙壮 SUN Zhuang(Qufu Normal University,Jining 272000,China)

机构地区：[1]曲阜师范大学,山东济宁272000

出　　处：《电子设计工程》2023年第23期41-44,49,共5页Electronic Design Engineering

摘　　要：信息化技术在相关领域应用逐渐加深,使得数据量急剧增加,造成数据存储空间不足、需求数据寻找困难等问题,制约着管理水平的提升与企业的可持续发展,因此构建基于稀疏表示的数据无失真压缩模型。应用超完备字典学习方法稀疏表示数据,以稀疏系数向量为基础,计算数据之间的相似度,基于谱聚类算法聚类处理数据,利用MTDimpute算法填补缺失数据,以缺失填补后的数据聚类集合为依据,通过PredZip算法(算术编码阶段与概率预测阶段)实现数据的无失真压缩。实验数据显示,应用构建模型获得的数据压缩增益最大值为18.8,压缩比最小值为0.1,压缩失真率最小值为0.5%,充分证实了构建模型数据压缩性能更优质,适合大力推广与应用。The gradual deepening of the application of information technology in related fields has led to a sharp increase in data,resulting in problems such as insufficient data storage space and difficulties in finding required data,which restrict the improvement of management level and the sustainable development of enterprises.Therefore,a research on the construction of data distortionless compression model based on sparse representation is proposed.Using the super complete dictionary learning method to sparsely represent data,based on the sparse coefficient vector,calculate the similarity between data,cluster the data based on a spectral clustering algorithm,and use the MTDimute algorithm to fill in missing data.Based on the data clustering set after filling in the missing data,use the PredZip algorithm(arithmetic coding stage and probability prediction stage)to achieve distortionless compression of the data.Experimental data show that the maximum data compression gain obtained by applying the construction model is 18.8,the minimum compression ratio is 0.1,and the minimum compression distortion rate is 0.5%.This fully confirms that the construction model has better data compression performance and is suitable for vigorously promoting and applying.

关 键 词：无失真压缩 数据预处理 稀疏表示 数据聚类 数据压缩 

分 类 号：TN01[电子电信—物理电子学]