Adaptive numerical smoothing: an efficient method of conditioning physiological signals

The main aim of the paper is to present a method of reducing the number of necessary operations in the process of numerically smoothing digitised signals, together with increasing the quality of smoothing, using suggestions which are new in the technique of numerical smoothing. The first suggestion consists of an analytical description of the smoothing process and the representation of the input mixture and undergoing smoothing signals by a series of sinc functions. The second proposal consists of using adaptive smoothing, i.e. smoothing with varying step and with some special means for increasing the quality of smoothing. These procedures must be calculated together with the processing and this procedure causes a small delay in the output signal. The approach presented is a deterministic one, i.e. the information about all instantaneous values of the input mixture is used instead of an a priori average information about disturbances. Attention is directed towards disturbances which are associated with physiological signals such as pressure, flow and electrocardiogram waveforms. The paper contains the basic theory and some experimental results which demonstrate the potential of adaptive smoothing.