Multimedia data are extremely valuable due to the amount of information they contain. At the same time, extracting information about the content of such data is very useful, yet very difficult and demanding in terms of computational time.
Extracting information is one of the main axes of our research. It is covered within two research programs – Soft Computing Methods with Applications for Supercomputers, and Recognition and Presentation of the Information from Multimedia Data.
The theme includes presenting the information that exists in the data, specialised algorithms for processing speech, images, video, 3D data, as well as structural and spatial data. With regards to data mining, formal languages and formal grammar is used, along with statistical techniques trained on huge data volumes and semantic web technologies.
This area also includes special mathematical methods that allow the use of inaccurate information. Such methods may be used when creating system models that are very close to reality. These models, also known as "fuzzy models", offer the following opportunities:
- Control complex systems,
- Make decisions with the use of multiple criteria that are not always quantifiable,
- Optimise the bases of not always precisely specified information,
- Analyse and prognosticate timelines (LFL Forecaster software is used),
- Process images (compression, fusion, edge detection, reduction, etc.),
- Recognise deformed symbols (letters, numerals),
- Filter signals,
- Search data for knowledge (LFL Miner software is used),
- Solve differential equations, etc.