Media signal processing

Research focus on processing of media information such as audio, video, and documents (e.g. text and computer data).

For audio, fundamental research on adaptive array signal processing underpins systems-oriented research on transparent audio communication and audio signal processing. This work deals with complex, non-stationary, multi-sensor systems, and is supported by STW, BTS and industrial contracts. Fundamental work emphasizes adaptive systems and filter banks, and algorithmic complexity reductions and performance improvements via block processing, blind source separation, and time-frequency trade-offs. Systems work concerns real-time implementation of complex adaptive schemes for e.g. source localization and separation, for suppression of artefacts such as reverberation and acoustic echoes, and for semantic audio analysis. A dedicated audio laboratory supports algorithmic studies and real-time experiments.

For video, fundamental research on depth image compression and on image/video segmentation, object-oriented compression and enhancement underpins systems-oriented research on video coding and architectures. This work pursues state-of-the-art video functionality at high cost-efficiency (e.g. for real-time aspects and/or mobile applications), through a combination of novel algorithms and architectures. Key applications include real-time face detection and recognition, scalable video coding for mobile computing, video segmentation, and image enhancement after digital compression. Detailed information about the The Video Coding and Architectures research sub-group can be found here.

Extensive communication and travel arrangements ensure that both sites together form a virtual ‘one-roof’ team. For documents, fundamental research on universal source coding based on context-tree weighting lends support to efficient implementation of record-breaking data-compaction techniques. The close connection between universal compression and learning is currently a hot research topic worldwide. Our research focuses on the applicability and performance of context-based compression schemes for learning in discrete and continuous spaces. Flows of ideas across these modalities are actively pursued, contributing e.g. to novel lossless audio-compression, semantic audio analysis, and digital watermarking schemes.

Meer weten?

Wil je meer informatie over de inhoud van het dossier of in contact worden gebracht met de kennisaanbieder? Neem dan contact met ons op.


Terug naar overzicht »