Estimation of Asymmetry in Facial Actions for the Analysis of Motion Dysfunction Due to Paralysis

A mesh-based approach to estimate asymmetries in facial actions for the analysis of facial motion dysfunction due to paralysis is presented. Pre-designed facial actions are captured using a laser rangefinder system and then analyzed in three-dimension for patients with Bell's palsy. The asymmetry computations are based on analyzing the deformations in a symmetric generic face mesh that was adapted to the range maps of measured facial actions. In the mesh adaptation, 42 nodes are pre-determined as mapping points and extracted from the corresponding color image, which was taken at the same time as shape measurement. These mapping points are used to calculate the coefficients of a polynomial function, which is approximated by the least squares estimation during rest of the node mapping. A 3D mesh model of each action is thus generated by iterative adaptation to the range data. The asymmetry is computed by analyzing the amount of deformation in identical patches of left and right sides of the 3D mesh of each action against those of relaxed condition. Asymmetry distributions are obtained by plotting right patch deformations against the left. Standard error and standard deviation of distributions of each patient are obtained and analyzed to determine the degree of motion dysfunction.