Human-centric design of intelligent systems

Our group’s activities are focused on the design of intelligent systems that are not black boxes, but rather designed for maximum reliability, responsibility and transparency in human use.

A central step towards this goal is achieving optimal and interpretable information integration in machine perception. We are therefore working on a tighter coupling of reliability-aware sensing, statistical signal processing, and machine learning.

On the applications side, we are using audiovisual input for highly robust speech recognition, e.g. for speech training and logopedics, reliable and interpretable human-machine interaction, and for speech quality enhancement and speech intelligibility optimization.

Smart sensing, signal processing and machine learning are also used as tools for fitness monitoring, sports analytics, technical diagnostics and automatic fault monitoring.