Facial Paralysis Dataset

Facial paralysis is the loss of facial muscle movementeither in one side or both sides of the face due to the facialnerve damage. Currently, the subjective assessments are widelyused techniques to determine the measure of degree with whichthe patient is affected. However, the subjective assessments arehighly dependant on the expert’s view and a few sets of gradingrules. In this paper, the quantitative assessment to measure thedegree of facial paralysis is proposed. The video database offacially paralyzed patients, which consists of seven different viewsand multiple subjects with ten different expressions are collectedunder three experts supervision. Inorder to capture the variationspresent in multiple views and subjects across all the expressions, alarge Gaussian mixture model (GMM) is trained. A feature vectoris obtained from each expression using a maximum a posterioriadaptation (MAP). The dimension of the adapted feature vectoris very high and contains redundant attributes. So, we reducethe dimension of the feature vector using factor analysis, whichcontains pertinent attributes. The proposed work is evaluated onthe video database of 39 facially paralyzed patients of differentage groups and gender. Based on the facial paralysis effect,experts assign subjective scores to patients using Yanagiharagrading rules, which are further used as ground truth. We alsoshow the efficacy of the proposed approach by measuring thedifferent degree of facial paralysis for all 10 types of expressionsbetter than existing approaches for quantitative assessment.