摘要
针对基于范数优化的多目标动态荧光分子重建稀疏性不足、定位精度低的问题,本研究提出了基于块稀疏贝叶斯学习的重建方法。该方法利用多观测动态荧光信号共享相同稀疏结构的特点,充分挖掘了信号的时空相关性信息和分块稀疏特性,将多观测模型联合稀疏重构问题转化为块稀疏贝叶斯学习问题,通过稀疏贝叶斯框架下的相关向量机稀疏学习模型获取稀疏解。实验结果表明,本研究算法相较于传统压缩感知算法具有更高的稀疏度和定位精度。
Aiming at the problem of insufficient sparsity and low positioning accuracy of multi-objective dynamic fluorescence molecular reconstruction based on norm optimization,a reconstruction method based on block sparse Bayesian learning was proposed.The characteristics of multi-observation dynamic fluorescence signals was utilized to share the same sparse structure,the spatiotemporal correlation information and the sparseness of the block were fully exploited,and the multi-observation model and the sparse reconstruction problem into block sparse Bayesian learning problems were transformed.The sparse learning model of relevance vector machine under sparse Bayesian framework was adopted to obtain sparse solutions.Experimental results show that the proposed algorithm has higher sparsity and positioning accuracy than the traditional compressed sensing algorithm.
作者
臧青杨
陈春晓
杨俊豪
李东升
ZANG Qingyang;CHEN Chunxiao;YANG Junhao;LI Dongsheng(Department of Biomedical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China)
出处
《生物医学工程研究》
2018年第4期454-459,464,共7页
Journal Of Biomedical Engineering Research
关键词
稀疏贝叶斯学习
压缩感知
动态荧光
三维重建
多观测向量
块稀疏模型
Sparse Bayesian learning
Compression perception
Dynamic fluorescent molecule
Three-dimensional reconstruction
Multiple measurement vectors
Block sparse model
