Ongoing projects
Concluded projects

Publications
Ph. D. theses
M. Sc. theses

Connections

Within the Laboratory research there were more than 100 concluded research and development projects in the past. Following is the list of titles and links of more important projects recently concluded:

* SLO-HRV Project 17/01-03

Multi-Agent Systems for Interpretation of Dynamical Scenes (N. Pavešić principal investigator for Slovenia)
The aim of the project is to provide a new model of a multiagent system for real time dynamic scene analysis based on hierarchical organization of agents with different capabilities (e.g. perception, transformation from iconic to symbolic world representation, etc.). A special attention is directed to models for temporal and fuzzy temporal knowledge representation and reasoning.

* SLO-HRV Project 23/01-03

Bilingual Speech Data Base (F. Mihelič principal investigator for Slovenia)
The aim of the project is to develop and provide bilingual speech data base for Croatian and Slovene speech. Speech data base is necessary for development of intelligent computer systems, capable of recognition and understanding Croatian and Slovene speech. Speech and text corpora for both languages are useful also for research, using computer technologies, on the fields of phonetics, morphology, lexicography and general linguistics. Database will provide speech and text corpora of both languages following the international standards used in speech technologies domain.

* Program MŠZŠ code P0-0517-1538 , 1.1.99 - 31.12.03;

Pattern Recognition (research group members: N. Pavešić, A. Babić, F. Mihelič, J. Gros, S. Dobrišek)
Our research goal is to focus both on basic and applied research that includes analysis of a) visual patterns, b) speech patterns, c) biomedical patterns, and d) environmental patterns.

a) This research topic concerns robot vision, active vision, pose and shape determination, appearance and multiresolution object representation. Distinguished and important research issues are next view planning in relation to the defined task, eye-arm coordination during reaching and grasping objects within the non-calibrated, or weakly-calibrated robot vision systems.

b) We concentrate on the language dependant procedures and factors of speech analysis and recognition. That in particular concerns: collecting and documenting Slovene language speech databases and text corpora, determining and selection of suitable basic speech units for speech recognition and synthesis, assembly of specialized vocabularies and lexicons of Slovene language. We can point out additional important research tasks such as development of tools for recognition of various, and for Slovene language, specific acoustic and prosodic measurements and their implementation, as well as determining and developing subsystems for syntactic and semantic decomposition of Slovene speech.

c) We are focusing on analysis and recognition of biomedical patterns for diagnostic purposes. Using procedures of machine learning we are going to build knowledge database needed to establish and follow-up diseases. Knowledge database, extracted from biomedical data samples, will be used in the system for automatic diagnosing and development of educational multimedia system for both medical students and physicians.

d) The intention is to analyze pollution of rivers and water flows, pollution of air and soil that can be done based on environmental data samples and patterns such as satellite data, patterns from automated measurement stations and manually collected samples.

* HP Voice Web Initiative ; (principal investigator F. Mihelič)

The aim of the project is to develop an experimental version of the voice driven text-to-speech system devoted for blind or visually impaired persons for reading Slovenian texts. Users can obtain texts from the Internet site of the Association of Slovenian Blind and Visually Impaired Persons Societies from their Electronic Information System (EIS) where users can find daily newspapers, some novels and other information. The system consists of four main modules. Web page demonstrating system capabilities (synthesis, recognition) will be also public available at the end of the project.

* Project MŠZŠ in ZDSSS code L2-2109 , 1.1.00 - 30.6.02

Automatic intelligent Slovene text reader for blind and visually impaired people (principal investigator N. Pavešić)
The aim of the proposed project is to develop an user-interface prototype, capable of reading arbitrary Slovene input texts. The sytem will be designed in a user-friendly manner so as to facilitate its usage for blind and visually impaired poeple. The user will be able to access information on the internet by giving spoken commands, the answers of the system and the contents of the internet pages will be reproduced by a synthetic voice.

* Project MŠZŠ code J2-2096 (C), 1.1.00 - 30.6.02;

Study of naturalness of synthesized speech (principal investigator J. Gros)
The aim of the project is to develop a text-to-speech system capable of producing high quality, intelligible, near to natural Slovene speech. We expect to achieve the required naturalness of the synthetic speech by introducing a corpus-based selection of basic speech units for concatenative speech synthesis. Further, the existing prosody modeling models have to be redefined and improved on real speech data, especially supra-segmental modelling.