DECONVOLUTION-BASED PHYSIOLOGICAL SIGNAL SIMPLFICATION FOR PERIODICAL PARAMETER ESTIMATION

The estimation of physiological parameters from raw sensor signals is absolutely crucial in modern clinical applications. A wide variety of these parameters incorporate a periodic nature, such as the heart rate or the respiration rate. This property can be exploited for their estimation. Particularly challenging is the processing of novel, unobtrusive measurement techniques, which are characterized by complex, time-varying waveforms. Simple peak detection algorithms are often not suited for these applications. One way to tackle these challenges is a preprocessing step for the simplification of the physiological signals. A novel deconvolution based approach for this preprocessing is introduced and evaluated in this paper. Two deconvolution methods are regarded, the “Minimum Entropy Deconvolution” (MED) and the “Maximum Correlated Kurtosis Deconvolution” (MCKD). Important parameters are outlined and examined. Finally, the methods are validated using artificial as well as real clinical signals to demonstrate their potential