Abstract
Maximum and accurate information of blood components are expected by clinicians during medical sample detection and disease monitoring. Photoplethysmography (PPG) and its morphological features are commonly used for measurement and analysis but significant fundamental limitations including spectral noise, interaction of subjects' variability portions (skin, fat and muscle) and the absorbing components (water, and blood) still affects the quality of the PPG morphology in optical based extractions of blood components and spectral monitoring. Dynamic spectrum (DS) is a potential clinical method to consider these effect in the physiological PPG sample. In this article, we proposed a novel filtering DS based method using adaptive time varying filter (ATVF) to consider the effect and meaning of spectral amplification of the physiological PPG signal. The spectrum PPG sample was partitioned into two equal halves of upper and lower (UL) sample to ensure spectral amplification and analysis with DS. The results from 184 clinical volunteers confirms that spectrum PPG signal amplification significantly influences the accuracy of the noninvasive optical analysis of blood components by adding noise effect due to signal weakening but increases the spectral component. A radial basis function (RBF) neural network modeling of the calibration and prediction data sets shows the performance of the proposed ATVF-UL extraction by reducing significantly the induced noise effect for smooth DS component and maximum detection of blood components.