Current Research Topics - Herbert Buchner,
Machine Learning for Signal Processing

The main research interests are currently centered around future human-machine interfaces of (networked) multimedia systems and data analysis/scene analysis methods, in particular

The common aim is to provide seamless, undistorted, and intuitive communication between users / interaction between users, the environment, and complex devices, e.g., infotainment systems or biomedical devices.

In order to achieve this we work primarily with multiple sensors and/or actuators, i.e., transducer arrays and sensor networks. To find optimum solutions for the related signal processing and machine learning tasks, various constraints depending on the particular application should be taken into account (e.g., whether there is reference information/reference signals or not leading to supervised vs. blind/semi-blind adaptive processing). Particular excitements of the above mentioned applications acoustic application lie in the broadband character of the signals and the time-varying, convolutive (multi-path and possibly dispersive) nature of the physical systems. From the theoretical point of view one of our major goals is thus to rigorously develop a novel unified framework for broadband multichannel communication and data acquisition based on first principles of information theory, estimation theory, and wave physics. Several of the intermediate results can be found in the related publications. It can be shown that both many of the well-known algorithms in the literature but also numerous novel and efficient algorithms can be derived as special cases of this generic framework. Some of these novel algorithms have already led to efficient real-time systems.

Prominent example: speech and audio processing for communication acoustics Important foundations for these topics include
  • Statistics

  • statistical signal processing

  • Machine learning

  • Efficient algorithms for broadband adaptive MIMO filtering

  • Pattern classification and analysis

  • Audio/video/biomedical sensing technologies and sensor fusion

  • Real-time multichannel signal processing, especially for speech and audio (Linux, Windows,...)

  • Wave physics

  • Psychoacoustics, binaural hearing

Acoustic human-machine interface

Example: contribution of the acoustic hands-free interface for EMBASSI, a former research project funded by German Federal Ministry of Education and Research (BMBF), which led to the first realtime full-duplex system for surround sound reproduction back in 2000/2001. At the 'Internationale Funkausstellung' (IFA) 2001 in Berlin, we presented the first Smart TV with hands-free voice control and gesture recognition.

Acoustic human-machine interface

Back to Herbert Buchner Back to the homepage of Herbert Buchner
Last modified: Jun 2011