Facial Emotional Expression Assessment in Parkinson's Disease by Automated Algorithm Based on Action Units

This work is devoted to the study of expression and interpretation of six basic emotions: anger, disgust, fear, happiness, sadness, surprise in patients with Parkinson's disease in comparison with the healthy control group of patients. The study involved 16 patients in each group. Each patients face was recorded using a 2D camera while performing 3 tasks: displaying a neutral state, displaying 6 basic emotions by researcher request, displaying 6 basic emotions depicted on the images. Action units were determined on each video frame. The percentages of emotional expressions in each video were determined, and the intensity of the recognized expressions for each task using the emotion recognition algorithm based on action units. The difference between emotional expressions and the neutral state was calculated as Euclidian distance between vectors of action units to quantify the changes in facial expression between the Parkinson's disease and healthy control groups. To analyze the differences between the groups, the non-parametric MannWhitney U-test was used. The obtained results show changes in the emotional expressions in the Parkinson's disease group in comparison with the healthy control group, Parkinson's disease patients show a decrease in the expressiveness of face and the intensity of the emotional expression.