A Template Addition Method for Eigentriple Rearrangement in Singular Spectrum Analysis for Processing Biopotential Signals With Extremely Lower SNRs

Biopotential signals captured by wearable devices are generally corrupted by different kinds of noises. Singular Spectrum Analysis (SSA) algorithm can be utilized for processing such signals by reconstructing the signal from the sum of a small number of independent and interpretable components. However, in case of extremely lower signal-to-ratio (SNR) signals, the components to be extracted are often located at the back of the eigentriple list. By analyzing the filtering characteristics of the SSA algorithm, a target component advancing method based on template addition is proposed. The template is constructed based on the frequency features of the processed signal. By splicing the template at the starting point of the processed signal, the energy of the components to be extracted is enhanced and therefore the eigentriples of interest can be effectively advanced to the front of the list. This method is applied to denoising and feature extraction of human physiologic signals with lower SNRs, such as non-contact Electrocardiogram (ECG), Ballistocardiogram (BCG) and Electroencephalogram (EEG). Experimental results show that, with the aid of the proposed method, the computational cost of the eigentriple selection in SSA can be effectively reduced.