Research program
Advanced methods of interaction in telecommunication
Code: P2-0069 (B) – included in ARRS records
Period: 1.1.2018 – 31.12.2023
Range in 2019: 2,67 FTE
Head: Kačič Zdravko
Link: www.sicris.si
Research program
In the field of multimodal interfaces, research is focused on developing new approaches of verbal and non-verbal communication, taking into account the target application domains of supportive living environments, and intelligent environments.
In the field of verbal communication technologies, research is focused on the development of new speech signal preprocessing processes, specifically acoustic signal segmentation into different categories, such as speech/non-speech, speech/music, speaker gender segmentation, and segmentation based on bandwidth of the channels. The results of this kind of research make it possible to implement speech recognition systems in real environments, where other acoustic signals appear along with the speech signal.
In the field of acoustic modeling, research is focused on defining procedures that will enable multimodal interfaces to be implemented on the basis of speech signal information (viseme recognition), enabling user-friendly communication with embedded virtual agents capable of verbal and non-verbal communication. In the field of language modeling, research is aimed at defining procedures that will enable the successful recognition of fluent and spontaneous Slovene speech, especially by using factorized language models with syntactic information, and language models with classes.
Also, research in this area involves the development of automatic machine text translation procedures that take into account the features of speech-oriented machine translation using statistical language models and factorized language models. The aim of the mentioned research is to develop a system of automatic recognition of fluent and spontaneous speech with a large dictionary of words for the Slovenian language.
The use of advanced speech segmentation techniques, acoustic modeling, and the use of factorized language models make it possible to achieve a significant increase in speech recognition performance for the test target Broadcast News application domain. Research also includes the design and construction of new language resources and the analysis of spoken discourse.
Within the research part for ensuring the correctness of systems, the research has been carried out in the field of formal specification and verification of modern systems. The research includes the development of a systematic approach to the formal specification of fair I/O automata with TLA, and the development of generic modules for use in verifying the various systems presented with these automata, with TLA+ and TLC model for verification of models. They also include research into the use of formal methods for determining effectiveness.
In the scope of this research they are developing an integrated SpinRCP development environment used for the formal verification of systems using the Spin.
Final reports on the results of the research program
PDF dokument (raziskovalni_program/ Napredne_URN_NBN_SI_DOC-9FJABTPD.pdf)
Zdravko Kačič
Head of the research program
E-mail: zdravko.kacic@um.si
Phone: +386 2 220 72 20
Office: G2-M18
Laboratories involved
Laboratory DSPLAB
As part of the research program, DSPLab deals with the field of speech technology and research in algorithms and methods that are world-class native languages, appropriately improving speech recognition and its role in various categories of telecommunications services, such as: intelligent environment, content capture and processing, natural communication.
Laboratory LEIS
Research in the domain of brain-computer interfaces is coming to the foreground, especially in the field of non-invasive methods of multichannel electroencephalographic signals acquisition and monitoring of rhythmic and less rhythmic cortical fluctuations to create a communication channel or indirect management of user environments, computer games, and support for disabled people.
Laboratory for Microcomputer Systems
In the area of ensuring the correctness of systems, both action-based and status-based approaches have their own good and bad features. The current approach to the specification and reasoning of reversible systems represents an alternative to the prevailing action-based approaches to date.
Program activities
