Abstract-Over the last 20 years, several different techniques have been proposed for computer recognition of human faces. The purpose of this paper is to compare two simple but general strategies on a common database (frontal images of faces of 47 people: 26 males and 21 females, four images per person). We have developed and implemented two new algorithms; the first one is based on the computation of a set of geometrical features, such as nose width and length, mouth position, and chin shape, and the second one is based on almost-grey-level template matching. The results obtained on the testing sets (about 90 % correct recognition using geometrical features and perfect recognition using template matching) favor our implementation of the template-matching approach. Index Terms-Classification, face recognition, Karhunen-Loeve expansion, template matching.

A person recognition system that makes use of acoustic and visual features is described. The system combines features at a score level and is capable of performing either an identification or a rejection. The improved performance of the integrated system with respect to the separate subsystems (acoustic and visual) is quantified. Keywords--- Normalization, Integration, Rejection. 1. Introduction Person identification and verification is a crucial task performed by both people and automatic systems in every day life. Since in many situations mistakes must be avoided (for example, banking transactions, database inquiry, access control to reserved areas or buildings and so on) it is necessary to adopt secure and reliable techniques to assure the desired degree of safety. The approach chosen at IRST to cope with this problem [1] is the integration of different types of distinctive cues of the person. A system that uses both acoustic features, extracted from the user's speech signal, and ...

In this work we investigate the use of a speaker adaptation technique, for speech recognition, based on neural network spectral mapping. Different multilayer perceptron neural network architectures are analyzed in order to optimize the spectral difference reduction in acoustic data of two speakers. Experiments are carried out in a telecontrol environment, used to provide voice commands to a mobile robot, based on DTW pattern matching. Keywords: Neural Network, Parameter Optimization, Speaker Adaptation, Spectral Mapping, Speech Recognition. Corresponding author: Dr. Salvatore Rampone, Facolt`a di Scienze, Universit `a del Sannio - Benevento, Via Port'Arsa 11, I-82100 Benevento, Italy. Tel/Fax +39 0824 23013, E-mail: rampon@tin.it 1 1

this paper different algorithms for training multi-layer perceptron architecture are applied to a significant discrimination task in High Energy Physics. The OneStep Secant technique is compared with On-Line Backpropagation, the "Bold Driver" batch version and Conjugate Gradient methods. In addition, a new algorithm (Affine Shaker) is proposed that uses stochastic search based on function values and affine transformations of the local search region. Although the Affine Shaker requires more CPU time to reach the maximum generalization, the technique can be interesting for special-purpose VLSI implementations and for non-differentiable functions