The Cognitive Engineering group develops applications based on Computer-Vision, Machine-Learning and Computer-Graphics in combination with methods that model Human cognitive processes. Highly controllable and realistic settings, real-world sensor data and simulations, offer the opportunity for advanced experiments and at the same time a framework to design and optimize processes in Human-Machine-Interfaces. Our approach ultimately opens a new window to modern industries such as entertainment computing (games), communication research (information transfer, multimedia), medical systems, and personal assistance systems (automotive safety).
For example, we develop realistic computer-generated models that offer as research tool new opportunities to study human cognitive processes in interactive situations. Results from such experiments are in turn used to build effective Human-Machine-Interfaces. We mainly develop and employ Computer-Vision and Machine-Learning algorithms to quantify human performance in natural interactive situations. The results from perception experiments provide constraints for designing artificial vision algorithms and thus lead to more efficient approaches in, for example, artificial environment understanding. Realistic virtual worlds and animations, generated with Computer-Graphics, support us with "ground truth" to prototype artificial recognition algorithms and to determine the critical factors in human decision making. The semantic modeling of high-dimensional data spaces is another important goal of our work. It provides the basis to effectively study and optimize Human-Computer interaction.
Main research areas
Applied Computer Vision
We develop Computer Vision systems that can enhance and support the visual perception and decision making processes of humans. The components we investigate range from low-level image features to high level system components for, e.g., autonomous navigation or artificial scene interpretation.
We combine Computer Vision, Machine Learning and Computer Graphics with experimental methods of Cognitive Sciences in order to optimize technical interfaces between human and machine perception.
Automatic Movement Reconstruction
Technologies Rich movement models, for example dynamic morphable face models, have a large impact on the fields of computer vision, behavior monitoring, as well as computer graphics and animation. Attributes such as descriptiveness, semantics, and intuitive control are desirable properties. For robust applications we build models from high-quality and noisy depth data such as that produced by Time-of-Flight (ToF) cameras or devices such as Microsoft Kinect.
Signal Processing of Social Interactions
We can best understand human behavior in natural interactions. For example, we have developed a real-time facial expression control system. Major technological challenges are in general the reduction of noticeable feedback latencies without loosing accuracy. The inclusion of body motion capture allows to create animatable humans. Learning generative models of interactions is another important goal in order to understand complex behavior. Moreover, we want to understand the visual perception of social crowds, which in return will benefit the development of artificial vision systems in order to make complex predictions.