AI and Multimodal Analysis
An image, a location and a spoken sentence describe different parts of a situation. FILA investigates what information their combination adds and how that contribution can be tested. The focus is on concrete tasks: predicting movement and analysing usable land.
AURORA connects GPS histories with speech context; ReSeDiUm studies satellite imagery and observations from agricultural practice. The important question is how the information fits together in time, space and meaning. Evaluation also considers what can be recognised or predicted without the additional source.
GTA-NarrativeTraj connects coordinates and dialogue from Grand Theft Auto V in a controlled research dataset. The 2025 paper reports about 30 hours of gameplay recording and 8,392 non-empty utterances. Time-aligned movement and dialogue make it possible to study what language can reveal about the next waypoint.
The dataset and logging pipeline support experiments with time-aligned movement and speech data. Because the trajectories come from a game world, transfer to real road traffic requires separate evaluation. Explore the GTA-NarrativeTraj repository with data, source code, analysis notebooks and setup instructions. The dataset paper explains the construction and research questions.
Speech processing is also studied as a complete pipeline. A group publication compares noise suppression, voice activity detection and automatic transcription under defined noise conditions, making interactions between processing stages visible. Earlier VoicePredict work provides historical context; the project pages and publications explain the individual studies. Potential research and student topics can be discussed with the FILA team.
Selected publications
- Anastasiia Sapeha, Eduard Sariiev, Mykyta Sapeha, Ibrahim Kovan, Subashkumar Rajanayagam, Kirill Karpov, Maksim Gering, Dmitry Kachan, Eduard Siemens (2025). GTA-NarrativeTraj: Language-Aware Trajectory Prediction from GPS and Dialogue in an Open-World Simulator. Proceedings of the International Conference on Applied Innovations in IT (ICAIIT), 13(5), 193–199 DOI: 10.25673/122853.
- Anastasiia Sapeha, Ibrahim Kovan, Subashkumar Rajanayagam, Kirill Karpov, Maksim Gering, Dmitry Kachan, Eduard Siemens (2026). A Comparative Study of Noise Effects, VAD, and ASR Methods in Speech Recognition. Proceedings of the International Conference on Applied Innovations in IT (ICAIIT), 14(2), 103–110 DOI: 10.25673/123832.
Research Areas · Team and Contact
Explore theses and student projects at FILA, or discover topics for your own work.
