Classification of the Factors Influencing Center of Pressure Using Machine Learning and Wavelet Analysis

Abstract

Postural control is influenced by somatosensory, visual, and vestibular information. This study investigated the effects of open/closed eyes and external vibration stimuli on Achilles tendon on postural control in healthy adults. The visual and vibrational inputs were selected to induce somatosensory changes in quiet standing posture. Machine learning and wavelet analysis were used to classify the factors that influence postural control. Fifteen healthy subjects performed a quiet standing task under four different conditions: eyes closed, eye opened, with Achilles vibration, and without Achilles vibration. Results showed that both visual and vibration input conditions caused a significant difference in the mean velocity of anteroposterior direction in Center of Pressure (COP) shifting. This indicated that visual and somatosensory information changes have a significant role on postural control. The feasibility of using machine learning for the detection and classification of the factors influencing postural control function has been confirmed in this study. Combination of the features extracted by wavelet packet analysis and time-domain features showed to have the best classification outcome with F1-score of 96.77%.