Redundant Gaussian dictionary in compressed sensing for ambulatory photoplethysmography monitoring

Abstract

With the rapid development of wireless communication technology, wireless photoplethysmography sensors have emerged and been applied in clinics or daily healthcare. However, energy efficiency is still a major obstacle for such devices in long-term use. Data compression can reduce airtime over energy-hungry wireless links and improve the energy efficiency of the sensors. In this study, we explored a redundant dictionary-based compressed sensing scheme for ambulatory photoplethysmography data compression. We first proposed a redundant Gaussian dictionary, which was the column combination of circle shift atoms of Gaussian function. Besides, we came up with a simple method for dictionary generation. To demonstrate its efficiency improvement for compactly representing photoplethysmography signal, the proposed dictionary was compared with the dictionaries of discrete cosine transform, wavelet transform, Fourier transform, and Gabor transform in experiments using the database of IEEE Signal Processing Cup 2015. Results indicated the recovered photoplethysmography signals were essentially undistorted by the combination of redundant Gaussian dictionary, sparse binary measurement matrix, and smoothed l0 pseudo norm recovery algorithm. The average percentage of root-mean-square difference was 8.00%, and its ratio over 9% was 0.32 when the compression ratio was equal to 30%. Furthermore, after a 20 Hz anti-aliasing low-pass filtering, perfect full-band signals were recovered by using our method, where every frame satisfied the requirement of non-distortion diagnosis at 70% data reduction. The outstanding performance and low-complexity indicate the proposed redundant Gaussian dictionary is an excellent choice for using in compressed sensing enabled ambulatory photoplethysmography.