Department of Signal Processing and Machine Learning

The Department of Signal Processing and Machine Learning uses modern methods of machine learning to perform research in the field of speech processing, automotive image/sensor data analysis, and quality control in production processes.


In the field of speech processing we are mainly interested in methods of speech enhancement, noise reduction, acoustic echo cancellation, artificial bandwidth extension, in-car communication systems, as well as instrumental speech quality measures. Further topics are beamforming and standard-conformant enhancement of decoded speech and audio, emotion recognition, gaze detection. Two of our maybe most exciting basic research topics are automatic speech recognition and information fusion. Our application areas cover car and office communications, teleconference systems, hands-free systems, hearing aids, gateways, and mobile phone chipsets.


In the field of automotive image/sensor data analysis we contribute to automated driving on the level of environmental perception close to the sensor.  Our current focus are reliability in neural networks, network topologies for image segmentation, prediction and compression as well as methods for identifying corner cases in big data resources. Also fusion aspects play an important role.


In the field of quality control and production processes we are long-term active in automatic acoustics and image analysis (visual, NIR) for quality control during (online) or after production. Here we develop expert systems according to the big data principle, having access to data every day and thereby continuously improving both robustness and intelligence of the system.