Benefit of Multimodal Systems to Reduce Incorrect Classifications
Face recognition and ID-verification techniques need to be really robust when used outside the laboratory. Conventional identification methods use color or gray-scale images of the person to be identified. Performance and precision of these systems depend on illumination, head-holding or facial expressions. Involving three-dimensional information for the description and recognition of faces, robustness and accuracy of person identification can be improved. Even in the case of a desired change in the appearance, various facial characteristics can only be modified with high effort, such as plastic surgery.
This includes the inherent facial geometry, which can additionally be used for identification by the use of three-dimensional features. Possible solutions to reduce incorrect classifications, include 2.5D/3D data from depth-of-view cameras. These multimodal techniques combine the strengths of conventional 2D face recognition with additional information of 3D facial geometry. Improvement of performance generated through multimodal systems has already been proven in the literature several times.