基于过完备字典稀疏表示的多通道脑电信号压缩感知联合重构

该文基于多通道脑电信号时空特性构建非正交变换过完备字典，准确稀疏表示蕴含时空相关性信息的多通道脑电信号，提高基于时空稀疏贝叶斯学习模型的多通道脑电信号压缩感知联合重构算法性能。实验选用eegmmidb脑电数据库的多通道脑电信号验证所提算法有效性。结果表明，基于过完备字典稀疏表示的多通道脑电信号，能够为多通道脑电信号压缩感知重构算法提供更多的时空相关性信息，比传统多通道脑电信号压缩感知重构算法所得的信噪比值提高近12 dB，重构时间减少0.75 s，显著提高多通道脑电信号联合重构性能。

A New Joint Reconstruction Algorithm of Compressed Sensing for Multichannel EEG Signals Based on Over-complete Dictionary Approach

In this paper, the over-complete dictionary with nonorthogonal factor is firstly gained from Electro Encephalo Graph (EEG) signal with spatio-temporal characteristics, and then it is used to sparsely represent multichannel EEG signal for containing the information of spatio-temporal correlation. This contributes to enhance the performance of the joint reconstruction of multi-channel EEG signal using the Spatio-Temporal Sparse Bayesian Learning (STSBL) algorithm. The multi-channel EEG signal from the open eegmmidb database are selected to evaluate the effectiveness of the proposed algorithm. The experimental results show that the designed over-complete dictionary can provide more valuable information about the spatio-temporal characteristics in multichannel EEG signal for STSBL algorithm. When compared to the existing conventional compressed sensing technique for reconstruction multi-channel EEG signal, the signal-noise ratio of the proposed method increases by 12 dB and the reconstruction time decreases by 0.75 s, which significantly improve the performance of joint reconstruction of multichannel EEG signal.