心电压缩感知恢复先验块稀疏贝叶斯学习算法

发布时间：2018-03-30 06:59

本文选题：先验块稀疏贝叶斯学习　切入点：压缩感知　出处：《仪器仪表学报》2014年08期

【摘要】：压缩感知在低成本、低功耗、长时间的无线心电信号应用上具有优势。但现有重构算法中存在重构信号质量不理想、较大的计算量以及不能自适应噪声变化等问题。本文针对非稀疏心电信号快速精确压缩感知重构提出了先验块稀疏贝叶斯学习(P-BSBL)算法。算法在块稀疏贝叶斯学习基础上,根据心电信号先验引入了近似零解空间初值设置和数字特征迭代停止条件。为了验证算法效果,提出的方法在MIT-BIH心电数据库上进行了仿真实验。实验结果表明P-BSBL能够实现高效非稀疏心电信号高信号质量重构。P-BSBL在正常和非正常心电信号重构上都优于凸优化和贪婪方法;适用于高数据压缩比和噪声变化的心电信号重构。
[Abstract]:Compression sensing has advantages in low cost, low power consumption and long time application of radio ECG signals, but the quality of reconstructed signals is not ideal in existing reconstruction algorithms. In this paper, a priori block sparse Bayesian learning algorithm based on block sparse Bayesian learning is proposed for the fast and accurate compression sensing reconstruction of non-sparse ECG signals, which is based on block sparse Bayesian learning. According to the prior ECG signal, the initial value setting of approximate zero solution space and the iterative stopping condition of digital feature are introduced. In order to verify the effect of the algorithm, The proposed method is simulated on the MIT-BIH ECG database. The experimental results show that P-BSBL can achieve high signal quality reconstruction of high performance non-sparse ECG signal. P-BSBL is superior to convex optimization and greedy method in both normal and abnormal ECG signal reconstruction. It is suitable for ECG signal reconstruction with high data compression ratio and noise variation.
【作者单位】：东南大学仪器科学与工程学院;邓迪大学医学科学与技术研究所;
【基金】：国家自然科学基金(61240032) 江苏省自然科学基金(BK2012560) 江苏省研究生创新基金(CXZZ13_0089) 中央高校基本科研业务费资助项目
【分类号】：TN911.7

【相似文献】