Motion Artefact Removal using Single Beat Classification of Photoplethysmographic Signals

A Photoplethysmographic (PPG) signal classification scheme for motion artefact detection is proposed. In the feature extraction, the PPG signal is modeled as a periodic signal with fundamental frequency and harmonics using a least-squares based algorithm. The extracted features correctly represent the morphological information of the PPG signal. A support vector machine (SVM) is used in the classification. A procedure is adopted to make the feature set scale and position independent, which improves the classification accuracy. Using 5 harmonics, a classification accuracy of 96.55% is achieved. As 26 different normal PPG morphologies are used in the classifier, the system would be able to identify any normal PPG signal among motion artefacts on a single beat basis.