Text this: Reduction of Vision-Based Models for Fall Detection