当前位置: X-MOL 学术Signal Process. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Distributed Compressive Sensing via LSTM-Aided Sparse Bayesian Learning
Signal Processing ( IF 4.4 ) Pub Date : 2020-11-01 , DOI: 10.1016/j.sigpro.2020.107656
Haijian Zhang , Wusheng Zhang , Lei Yu , Guoan Bi

Abstract Model-driven algorithms on distributed compressive sensing with multiple measurement vectors (MMVs) have been generally based on the assumption that the vectors in the signal matrix are jointly sparse. However, the signal matrix in many practical scenarios violates the above assumption, since there might exist unknown dependency between vectors. It highlights the limitation of model-driven approaches and the necessity to move toward data-driven ones. In this paper, we propose a data-driven algorithm, which interprets the MMV problem as sequence modeling to infer the unknown dependency and encourage sparse signal recovery within the framework of sparse Bayesian learning (SBL). Specifically, we extend the fast-SBL algorithm suitable for MMV sparse recovery. Then the long short-term memory (LSTM) is introduced into the extended fast-SBL, serving as a strategy for selecting the basis function from the dictionary. Compared with existing data-driven algorithms, the proposed LSTM-SBL algorithm inheriting the characteristics of fast-SBL has fewer local minimums and better robustness to noise. Extensive experiments are conducted on MNIST and MSR datasets to evaluate the accuracy, robustness and speed of the LSTM-SBL algorithm. Experimental results are presented and analyzed to illustrate the potential advantages of the LSTM-SBL algorithm in contrast to state-of-the-art MMV recovery algorithms.

中文翻译:

通过 LSTM 辅助稀疏贝叶斯学习的分布式压缩感知

摘要 具有多个测量向量 (MMV) 的分布式压缩感知模型驱动算法通常基于信号矩阵中的向量联合稀疏的假设。然而,许多实际场景中的信号矩阵违反了上述假设,因为向量之间可能存在未知的依赖关系。它强调了模型驱动方法的局限性以及转向数据驱动方法的必要性。在本文中,我们提出了一种数据驱动算法,该算法将 MMV 问题解释为序列建模,以在稀疏贝叶斯学习 (SBL) 的框架内推断未知依赖性并鼓励稀疏信号恢复。具体来说,我们扩展了适用于 MMV 稀疏恢复的快速 SBL 算法。然后将长短期记忆(LSTM)引入扩展的fast-SBL,作为从字典中选择基函数的策略。与现有的数据驱动算法相比,所提出的LSTM-SBL算法继承了fast-SBL的特点,具有更少的局部最小值和更好的噪声鲁棒性。在 MNIST 和 MSR 数据集上进行了大量实验,以评估 LSTM-SBL 算法的准确性、鲁棒性和速度。提出并分析了实验结果,以说明 LSTM-SBL 算法与最先进的 MMV 恢复算法相比的潜在优势。在 MNIST 和 MSR 数据集上进行了大量实验,以评估 LSTM-SBL 算法的准确性、鲁棒性和速度。提出并分析了实验结果,以说明 LSTM-SBL 算法与最先进的 MMV 恢复算法相比的潜在优势。在 MNIST 和 MSR 数据集上进行了大量实验,以评估 LSTM-SBL 算法的准确性、鲁棒性和速度。提出并分析了实验结果,以说明 LSTM-SBL 算法与最先进的 MMV 恢复算法相比的潜在优势。
更新日期:2020-11-01
down
wechat
bug