3D Face Recognition Using Concurrent Neural Modules

We investigate 3D face recognition by proposing an algorithm with the following processing stages: (a) thresholding of depth maps of 3D range images; (b) normalization and alignment; c) feature extraction by Gabor Wavelet Filtering (GWF); d) Principal Component Analysis (PCA); e) classification using the concurrent neural model previously proposed by the first author called Concurrent Self-Organizing Maps (CSOM), representing a winner-takes-all collection of self-organizing neural network modules. For comparison to CSOM, we also evaluate the performances of several statistical classifiers (1-NN and K-Means). The implemented neural versus statistical classifiers are evaluated using GavabDB database containing 3D face images of 61 subjects. The best experimental result of CSOM leads to the recognition rate of 95.08 %, by comparison to the rate of 83.60 % obtained using k-Means and to that of 88.52 % given by NN. Key-Words: 3D face recognition, range images, Gabor Wavelet Filters, neural classifier, Concurrent SelfOrganizing Maps