Multitask Linear Discriminant Analysis for View Invariant Action Recognition