Projects of the speech analysis and synthesis department

Speech analysis and synthesis department is applying for grants using these grant schemes:VEGA – Scientific grant agency MŠVVaŠ SR and SAS (Vedecká grantová agentúra MŠVVaŠ SR a SAV)
GAV – Grant Agency for Science (Grantová agentúra pre vedu)
COPERNICUS – European Communities
ASFEU – MŠVVaŠ SR Agency for EU Structural Funds (Agentúra MŠVVaŠ SR pre štrukturálne fondy EÚ)
APVV – Research and Development Support Agency (Agentúra na podporu výskumu a vývoja)
V4 – Visegrad Fund
AFAROB: Innovative Aphasia Therapy using Robotics and AI

The project explores the integration of social robots and Artificial Intelligence (LLM) into the treatment of aphasia—a language disorder caused by brain damage. Since access to speech therapy is often limited, the goal is to develop a discourse-level approach that simulates natural communication and increases therapy intensity.

Key Objectives and Methodology:

  • Therapy Design: Developing procedures focused on spontaneous speech.
  • Technology Testing: Validating the use of robots and language models first with healthy individuals, followed by a clinical group of persons with aphasia.

Impact: Increasing therapy availability and frequency, leading to improved functional communication and long-term quality of life for patients.

ALOIS: Automated Diagnostics of Cognitive Disorders Using Speech Analysis and Robotics
Project Overview: The ALOIS project builds upon research in early detection of neurodegenerative diseases, focusing on integrating advanced Natural Language Processing (NLP) and social robotics into clinical practice.
 
Methodology: The system utilizes artificial intelligence for the automated evaluation of a patient’s speech during standardized diagnostic tests. A key element is the involvement of a social robot that acts as a testing mediator. This approach helps reduce patient stress and ensures objective data collection in a controlled environment.
 
Key Deliverables:
  • Robotic Platform: Development and testing of a methodology where a social robot conducts a dialogue with the examinee to collect acoustic and linguistic data.
  • Diagnostic Models: Algorithms capable of identifying subtle speech deviations (e.g., hesitations, vocabulary loss, syntax changes) that indicate early stages of cognitive decline.
  • Expert Database: Expansion of speech corpora with specific recordings of human-robot interactions for further linguistic research.
  • Clinical Application: A tool designed to support physicians and clinical psychologists by accelerating the screening process and providing precise quantitative data.

Impact: The ALOIS project pushes the boundaries from passive app-based testing toward active technological assistance in medical environments, increasing the efficiency of Alzheimer’s disease diagnosis in its earliest stages.
EWA: Early Warning of Alzheimer’s Disease
Project Overview: An interdisciplinary research project focused on the early detection of Alzheimer’s disease, Parkinson’s disease, and other neurodegenerative disorders through artificial intelligence and speech analysis.
 
Methodology: The application utilizes speech (naming and describing images), while a specialized algorithm evaluates changes in vocabulary, reaction speed, articulation, and speech fluency.
 
Key Deliverables:
  • EWA Mobile App: A unique mobile application designed for testing from the comfort of home.
  • EWA-DB Speech Database: A specialized database containing recordings from over 1,600 speakers for further scientific research.
  • Speech Recognizer: Developed directly at our institute, ensuring high diagnostic accuracy.
Consortium of Partners: The project bridges corporate and scientific expertise – participating members include AXON PRO (coordinator), Centrum MEMORY n.o., the Faculty of Informatics at PEVŠ, the Faculty of Medicine at UPJŠ, and the Institute of Informatics SAS.
The full text of the research article is available in the journal Scientific Data.
 
Funding: This project was co-financed by the European Union from the European Regional Development Fund under the Integrated Infrastructure Operational Programme.
Research on Atypical Voice Interfaces

Current voice interfaces are primarily optimized for the general population and standard, calm communication. To enhance the robustness of both theoretical models and practical applications, it is essential to research non-standard and extreme cases of human-machine interaction.

Project Goals and Focus:

  • Optimization for Specific Groups: Researching communication with seniors, pre-pubescent children, and individuals with neurodegenerative diseases.
  • Emotional and Linguistic Inclusion: Analyzing affective communication (various emotional levels) and developing voice interaction possibilities in the Romani language.

Impact: Advancing cognitive models and expanding the accessibility of voice technologies for diverse target groups and communication contexts.

Innovative Technologies for Forest Fire Management

The project focuses on advanced research and development of selected technological concepts for an integrated forest fire management platform. This work builds upon the foundations originally established within the H2020 project SILVANUS.

Core Focus:

  • Innovation: Developing advanced approaches for integrated information platforms.

  • Continuity: Expanding and optimizing technologies for real-world fire management scenarios.