Saltar apartados

Research Project

Técnicas Interactivas y Adaptativas para Sistemas Automáticos de Reconocimiento Aprendizaje y Percepción

Reference: TIN2009-14205-C04-01
Funding: Ministerio de Ciencia e Innovación
Period: from 01/01/2010 to 31/12/2013
Project Leader: Micó Andrés, María Luisa
Members: Calera Rubio, Jorge
Gómez Ballester, Eva
Oncina Carratalá, Jose
Pastor Seva, Pedro
Requena Jiménez, Antonio
Rico Juan, Juan Ramón
Serrano Díaz-Carrasco, Aureo
Summary of the project:
Traditionally, Pattern Recognition systems have been designed to work in an automatic way,
taking the place of the manual human intervention. Based on this paradigm, multiple and diverse systems have been
implemented in different domains. We proposed a reorientation of the current research with a new paradigm
where the systems will not
perform in an automatic way but in an interactive fashion. The main reason for this is that automatic systems are not
free from errors and, being high quality results the principal objective, a kind of supervision is needed. On the other
hand, as time goes by, intrinsic interactive applications are more important and frequent. For example, information
retrieval applications or computer aided decision systems. The main objective of the present project is to make use of
the information provided by that supervision in order to train again the models and having the opportunity of correcting
the forthcoming errors. In that case, the efficiency of the system is measured as a combination of the effort provided by
the user during the learning task (usually the number of interactions needed to fulfil it) together with the time the system
needs to re-train their models and provide a new hypothesis. The interactive paradigm poses new challenges to the
Pattern Recognition research community: incremental models learning (dynamic algorithms), new strategies for
hypothesis generation, making use of the inner data consistency (unsupervised learning) and many more that will
appear during the development of the project. These new strategies will give rise to the development of new concepts
and techniques, as a result of applying this new paradigm on many different methodologies, classic as long as the most
recent ones in the filed of Pattern Recognition.
Another challenge of this project consist of particularizing the above ideas to several interesting application domains as
human behaviour analysis, text transcription, content-based image retrieval, handwriting recognition, surveillance,
access control and biometrics systems, materials inspection and stellar spectra classification.

Departament of Software and Computing Systems

University of Alicante, carretera San Vicente s/n

03690 San Vicente del Raspeig, Alicante (Spain)

Tel: (+34) 96 590 3772 | Fax: (+34) 96 590 9326

Follow us on:
[Twitter] [Facebook]