Multi Library Wavelet Neural Networks for 3D face recognition using 3D facial shape representation

This paper presents a new approach for 3D face modeling and recognition. Motivated by finding a representation that embodies a high power of discrimination between face classes, a new type of 3D shape descriptors is suggested. We have developed a fully automatic system which uses an alignment algorithm to register 3D facial scans. In addition, scalability in both time and space is achieved by converting 3D facial scans into compact wavelet metadata. Our system consists in two phases. The first phase is called enrolment composed of 3 steps: data processing, alignment and metadata generating. The metadata generating step is powered by the use of Multi Library Wavelet Neural Networks (MLWNN). The second phase is called Authentication it starts with the calculation of depth distances between a probe and gallery 3D face. A K-Nearest Neighbors (K-NN) technique is used for 3D face classification. The results of this contribution are more interesting, in comparison with some others works, in term of recognition rate using the GavabDB 3D facial database.