Novel Signal Noise Reduction Method through Cluster Analysis, Applied to Photoplethysmography

Physiological signals can ofen become contaminated by noise from a variety of origins. In this paper, an algorithm is described for the reduction of sporadic noise from a continuous periodic signal. The design can be used where a sample of a periodic signal is required, for example, when an average pulse is needed for pulse wave analysis and characterization. The algorithm is based on cluster analysis for selecting similar repetitions or pulses from a periodic single.Tis method selects individual pulses without noise, returns a clean pulse signal, and terminates when a sufciently clean and representative signal is received. The algorithm is designed to be sufciently compact to be implemented on a microcontroller embedded within a medical device. It has been validated through the removal of noise from an exemplar photoplethysmography (PPG) signal, showing increasing beneft as the noise contamination of the signal increases. The algorithm design is generalised to be applicable for a wide range of physiological (physical) signals.