Kurtosis Rejection-Based Person Identification from Photoplethysmography Signals Recorded During a Step Exercise

Authors : Tuğba Aydemir, Mehmet Şahin

Pages : 1825-1835

Doi:10.31466/kfbd.1492325

View : 25 | Download : 29

Publication Date : 2024-12-15

Article Type : Research Paper

Abstract :In addition to computer vision-based methods, physiological signal-based person identification (PI) applications have attracted great attention in recent years with various kinds of advantages. Because physical activities can significantly contaminate physiological signals, most of the PI models were proposed to acquire the considered signal during the resting state. In this study, we proposed a kurtosis rejection-based PI from photoplethysmography (PPG) signals recorded during a step exercise. In the preprocessing stage, we rejected the PPG trials, which have a kurtosis value greater than five and are labeled as non-PPG, from the classification process. Afterward, the features were extracted by the sequential forward mother wavelet selection method and classified using the k-nearest neighbor algorithm. We achieved the highest classification accuracy rate of 88.64% PI performance. The obtained results proved that the kurtosis rejection-based PPG signals recorded during the step exercise can be reliably used for PI.
Keywords : Kişi tanımlama, Basıklık, Fotopletismografi, Dalgacık seçimi, Sıralı ileri, Özellik çıkarımı, Sınıflandırma