Multiscale principal component analysis to denoise multichannel ECG signals

In this work, multiscale principal component analysis (MSPCA) is introduced for denoising of multichannel electrocardiogram (MECG) signals. Wavelet decomposition of MECG signals segments the clinical information content at different Wavelet subbands or scales. At subband levels or scales multivariate data matrix are formed using Wavelet coefficients extracted from the same scales of MECG signals. At each subband matrix or scales, PCA is applied for noise elimination. To retain essential diagnostic components, matrices comprising of lower order Wavelet subbands are processed with reduced set of principal component (PC). Qualitative performance is evaluated and quantitative performance of denoising effect is measured by input/output signal-to-noise ratio (SNR). Signal distortion measures are evaluated using percentage root mean square difference (PRD), Wavelet weighted PRD (WWPRD) andWavelet energy based diagnostic distortion measure (WEDD). The proposed algorithm is tested with database of CSE multilead measurement library. The results show significant improvement in denoising of MECG signals with the lowest PRD of 3.488 and high SNR improvement of 34.279 dB.