Soft Computing Techniques for Human-Computer Interaction

Soft Computing Techniques for Human-Computer Interaction

Oscar Déniz (Universidad de Castilla-La Mancha, Spain), Gloria Bueno (Universidad de Castilla-La Mancha, Spain), Modesto Castrillón (Instituto Universitario de Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería, Spain), Javier Lorenzo (Instituto Universitario de Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería, Spain), L. Antón (Instituto Universitario de Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería, Spain) and M. Hernández (Instituto Universitario de Sistemas Inteligentes y Aplicaciones Numéricas en Ingeniería, Spain)
Copyright: © 2010 |Pages: 15
DOI: 10.4018/978-1-61520-893-7.ch003
OnDemand PDF Download:
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Soft computing aims at using tricks or shortcuts that do not provide optimal solutions but useful approximations that can be computed at a reasonable cost. Such approximations often come in the form of heuristics and “rules of thumb.” Computer vision relies heavily on heuristics, being a simple example the detection of faces by detecting skin color. Another approach that may also be considered as heuristics is the use of inductive learning, where the idea is to emulate humans in the sense that achieving certain skills require gradual learning. Thus, we would not make an effort to articulate solutions as equations, rules or algorithms. The solution would instead be sought automatically by feeding the system with training examples that would allow it to classify new samples. This chapter describes two successful applications of such soft computing approaches in the field of human-computer interaction, showing how the clever use of heuristics and domain restrictions can help to find solutions for the most difficult problems in this field.
Chapter Preview
Top

Eyewear Selector

The first application is a hardware-software system, intended for use at optical shops, which allows individuals to test different models of spectacles in a sort of real-time video mirror. The optical market is nowadays saturated with an increasingly complex array of lenses, frames, coatings, tints, photochromic and polarizing treatments, etc. The number of clients can grow only if the selection process is shortened or automated. A number of deployed systems have already demonstrated that eyeglass selectors can increase sales and customer satisfaction (Morgan, 2004).

From a research viewpoint, such systems represent an interesting application of Computer Vision, Multimedia and Human-Computer Interaction. The Augmented Reality, see a survey in (Azuma, 1997), of watching ourselves and try different “virtual” spectacles can be achieved by combining computer vision and graphics. The Magic Lens and Magic Mirror systems, for example, use the ARTag toolkit (Fiala, 2004), which mixes live video and computer-generated graphics. People can wear cardboard patterns that ARTag can detect. The graphics are placed in the visible positions of the patterns. The system can work in real-time on general-purpose hardware, although people have to wear the cardboard. Another approach is taken in (Lepetit et al 2003), where virtual glasses and moustaches are added to live video. Although the system works with an impressive frame rate of 25Hz the user must start the tracker by reaching a position that is close to a generic triangle-based face model shown on the screen. The ARMirror is a kiosk-based entertainment setup that shows live video overlaying virtual hats, skulls, etc., see (Lyn et al, 2005). Only the face as a whole is tracked, however.

Complete Chapter List

Search this Book:
Reset