Estimation of Pain from Normalized Energy of IMFs of EMG Signal

Pain is one of the most complicated and subjective sensation. As of now, doctors need to verbally communicate with the subjects for assessment of pain. In many cases, proper assessment of pain is not possible, especially for infants, deaf and dumb subjects, unconscious subjects etc. It is reported that pain stimulus excites the sympathetic nervous system (SNS) that changes the activity of few bio-signals that are directly related to SNS activities and the facial expression as well. In this present work, EMG of zygomaticus muscle which is related to change in facial expression is analyzed. Considering non-stationary nature of the EMG signal stimulated through pain, Empirical Mode Decomposition method is applied on EMG for its data adaptive nature. Standard Logistic Regression classifier is used on selected IMFs to classify between ‘no pain’ and different pain condition. Gradual increase in classification accuracy is noticed with the increase of pain intensity.