Enhancing 3D Face Recognition by Combination of Voiceprint

This paper investigates the enhancement of identification performance when using voice classifier to help 3D face recognition. 3D face recognition is well known for its being superior to 2D due to the invariance in illumination, make-ups and pose. However, it is still challenged by expression variance. The partial ICP method we used for 3D face recognition could implicitly and dynamically extract the rigid parts of facial surface and be able to get much better performance than other methods in 3D face recognition under expression changes. This work serves to further improve the performance of recognition by combining a voiceprint classifier into partial ICP method. We implement 9 combination schemes, and experiments on database of 360 models with 40 subjects, 9 3D face scans with four different kinds of expression and 9 sessions of utterance for each subject, shows improvement of performance is very promising.