基于周期元分析的运动伪迹消除方法——以心电信号为例

Study on Motion Artifact Reduction Based on Periodic Component Analysis Using ECG as a Case

运动伪迹是动态生理信号的主要干扰源,而在穿戴式检测系统中,由于干电极的使用,干扰更为严重。本文根据运动伪迹的时域瞬态性和生理信号固有周期性的特点,研究了一种基于周期元分析的运动伪迹抑制方法。该方法首先运用多分辨率分解将单通道信号转化为多通道信号,再实施周期元分析。然后以实测心电(ECG)信号为例,与基于经验模态分解和自适应滤波的消除方法进行分析比较,发现周期元分析的优势明显,可更有效地分离正常生理信号和运动伪迹。作为一种时域分析方法,周期元分析可实现频谱混叠信号的分离,对受污染ECG信号的波形特征实现有效复原,并可推广到其他生理信号的处理中。

Motion artifacts are a main interference source of ambulatory physiology signals. The interference in wear- able detection systems is more serious because of using dry electrodes. On account of the instantaneity in motion arti- facts and periodicity in physiological signal, we presented a new method based on periodic component analysis for motion artifact reduction. The single channel signal is transformed into multi-channel signal with multi resolution a- nalysis, and then periodic component analysis can help us to separate the normal physiological signal from motion ar- tifacts. A case study in electrocardiogram （ECG） demonstrates that periodic component analysis is better than the empirical mode decomposition and adaptive filtering methods. Periodic component analysis as a time domain method can discriminate the signal with frequency aliasing, and recover the ECG waveform feature corrupted. This method can be easily extended to other physiological signal processing.