This paper presents some results on the possibilities offered by neural networks for human face recognition. In particular, two algorithms have been tested: learning vector quantization (LVQ) and multilayer perceptron (MLP). Two different approaches have been taken for each case, using as input data either preprocessed images (gray level or segmented), or geometrical features derived from a set of manually introduced landmarks. The preprocessing steps included resolution reduction and

more »

... on. For the geometrical features´ case, a Karhunen-Loeve expansion was used to extract features among the different possibilities offered by 14 landmark points. For the experiments, a database composed of 300 images was used. The pictures correspond to 10 frontal, inclined o rotated views from thirty male persons of similar age and race. If gray level images are used as input data, the experimental results show higher recognition rates for LVQ than for MLP (96.7% versus 83.3%). Applying a previous segmentation stage strongly decreases the recognition rates. For geometrical features, the situation is reversed: MLP yields better results than LVQ (93.3% versus 84.4%).