Running wavelet archetype: time–frequency ensemble averaging requiring no fiducial points

Abstract

A novel method of signal averaging within the wavelet transform time–frequency plane is presented. This running wavelet archetype technique uses the inherent scale-dependent periodicity of the wavelet function within the averaging process, thus removing the requirement for the identification of fiducial points which is often the main cause of errors in traditional methods. The technique is illustrated using a synthetic signal and in its application to the detection of a cardiac pulse component in a noisy photoplethysmographic signal.