基于隐马尔科夫树模型小波域压缩采样信号重构方法.pdfVIP

第 24 卷 第 4 期 电子测量与仪器学报 Vol. 24 No. 4 · 314 · JOURNAL OF ELECTRONIC MEASUREMENT AND INSTRUMENT 2010 年 4 月 DOI: 10.3724/SP.J.1187.2010.00314 基于隐马尔科夫树模型的小波域压缩 采样信号重构方法* 赵贻玖 王厚军 戴志坚 (电子科技大学自动化工程学院, 成都 611731) 摘 要: 压缩传感理论利用信号的稀疏性, 对其非自适应线性投影进行压缩采样, 通过最优化问题准确重构原始信号。 传统重构算法仅利用了信号的稀疏性, 而未对转换后的信号结构进行分析。提出了一种基于 4 状态的隐马尔科夫树模型的小 波域压缩采样信号的重构方法, 相对 2 状态的隐马尔科夫树模型, 该模型能够获取相邻尺度小波系数的更多相关特性, 通过 仿真结果表明, 该算法具有更高的重构精度。 关键词: 压缩采样；非自适应线性投影；小波变换；隐马尔科夫树模型 中图分类号: TN911.7 文献标识码: A 国家标准学科分类代码: 510.40 Compressive sampling signal reconstruction in wavelet-domain based on hidden Markov tree model Zhao Yijiu Wang Houjun Dai Zhijian (School of Automation Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China) Abstract: Compressed sensing theory enables the reconstruction of sparse signal from a small number of non-adaptives linear projections. Conventional reconstruction algorithm involves linear programming or greedy algo- rithms, these reconstruction techniques are generic and assume no particular structure in the signal aside from sparsity. The compressive sampling signal reconstruction in wavelet domain is inspired based on tow-state wavelet hidden Markov tree model. In this paper, we propose a four-state wavelet Hidden Markov Tree model, it can capture more in- terscales dependencies of wavelet coefficients between two neighboring scales, the simulation shows that it reconstruc- tion precision is improved. Keywords: compressive sampling; non-adaptives linear projection; wavelet transform; hidden markov tree model 1 引 言