General Information
AI4TRUST-RO project was funded by the Romanian Government through the Executive Agency for Higher Education, Research, Development and Innovation Funding (UEFISCDI), programme “Horizon Europe - Institutional Awards”, project ID: PN-IV-P8-8.1-PRE-HE-ORG-2023-0078, contract number 27PHE/ 2023.
The implementation team comprised some of the Speech and Dialogue research laboratory members in POLITEHNICA Bucharest: Horia Cucu (project director), Dan Oneață, Alexandru Caranica, Lucian Georgescu, Gabriel Pîrlogeanu, Octavian Pascu.
The project started in January 2024 and was finalized in December 2025.
Activities
The activities performed in the project were the following:
- A1.1 Design and implementation of modern methods for multimodal emotion recognition
- A1.2 Experience exchanges
- A1.3 Dissemination
- A2.1 Work visits for international partnerships
- A2.2 Organizing a scientific conference
- A2.3 Research, design and implementation of speech recognition methods in acoustically adverse environments
The activities were 100% completed and all institutional research and development objectives were achieved.
Deliverbales
The deliverables created within the project were the following:
- D1.1 Implementation and dissemination report
- D1.2 Project website
- D1.3 A scientific article
- D2.1 Implementation and dissemination report
- D2.2 A scientific article.
Results
The project results are primarily represented by the deliverables mentioned in the previous section and the scientific publications produced by the project team.
The working visits and experience exchanges of the POLITEHNICA Bucharest team were carried out in Thessaloniki (Greece), Copenhagen (Denmark), Geneva (Switzerland), Lisbon (Portugal) and Rotterdam (Netherlands). As a result of these visits and experience exchanges, research topics of common interest were identified and connections were made that are being exploited in order to propose research projects in international consortia. For example, during this period we are working, together with the partners in Thessaloniki, on a project proposal for the UNITE - European Digital Health Valleys call.
One of the project activities involved the organization of a scientific conference. Thus, with the support of this project, in October 2025, the Speech and Dialogue Research Laboratory, from POLITEHNICA Bucharest, organized, in collaboration with the Technical University of Cluj-Napoca, the Institute of Informatics of the Romanian Academy and the Institute for Artificial Intelligence of the Romanian Academy, the international conference “Speech Technology and Human-Computer Dialogue” (SpeD). The aim of the event was to bring together researchers, engineers, developers and industry professionals to exchange innovative ideas and deepen research in cutting-edge areas: automatic speech recognition, voice synthesis, speaker identification, emotion recognition in speech, but also aspects related to audio security (deepfakes) and human-machine dialogue.
The conference provided remarkable academic visibility to Romanian research laboratories in this field. By indexing the papers in IEEE Xplore and Web of Science, SpeD will generate scientific impact and potential for long-term collaboration with institutions and companies globally.
The research activity carried out within the project led to the design and implementation of a method for speech recognition and processing in adverse acoustic conditions, a method for detecting counterfeit audio materials, a multimodal method for identifying emotions in speech and a method for diarizing speech/
The dissemination of the project results was carried out through the project website and by publishing several scientific articles. In total, four scientific articles were published: two at the most important speech processing conference in the world, Interspeech 2024, a conference already indexed by IEEE and ISI Web of Science, and two others at IEEE conferences, Communications 2024 and SpeD 2025.
Publications
- Pîrlogeanu, Gabriel, Alexandru-Lucian Georgescu, and Horia Cucu. "Open Source State-Of-the-Art Solution for Romanian Speech Recognition." In 2025 International Conference on Speech Technology and Human-Computer Dialogue (SpeD), pp. 102-107. IEEE, 2025.
- Georgescu, Alexandru-Lucian, Gheorghe-Iulian Chivu, and Horia Cucu. "Exploring Fusion Techniques for Multimodal Emotion Recognition." In 2024 15th International Conference on Communications (COMM), pp. 1-6. IEEE, 2024.
- Pascu, Octavian, Adriana Stan, Dan Oneață, Elisabeta Oneata, and Horia Cucu. "Towards generalisable and calibrated audio deepfake detection with self-supervised representations." In Proc. Interspeech 2024, pp. 4828-4832. 2024. ISI WOS:001331850104187
- Pîrlogeanu, Gabriel, Octavian Pascu, Alexandru-Lucian Georgescu, and Horia Cucu. "Hybrid-Diarization System with Overlap Post-Processing for the DISPLACE 2024 Challenge." Proc. Interspeech 2024, 1625-1629. 2024. ISI WOS:001331850101154