25 October 2021
Colloquium by Prof. Dr. Björn Schuller
Computer Audition (CA) is all about making machines listen – and understanding what they “hear”. The last decade or less has seen the field mature with “boost on”. This comes, as like for most if not any other pattern recognition task, deep learning and the general trends in AI have revolutionized the field. This came at massive improvements in performance and robustness. Yet, at the same time, increased application by the general public has brought along also ethical, legal, and social implications.
So, while “things” such as audio source separation, speech recognition, or speaker analysis and many more in CA are working considerably better today, they still need improvement for many use-cases. In parallel, a number of new concerns arose that urgently demand for technical responses.
Likewise, the presenter’s bucket list on rendering computer audition a fully satisfactory experience has recently grown. Beyond improving robustness and performance by self-learning of representations, coupling analysis and synthesis, and learning efficiently, ideally reinforced at scale, new items are presented. These include explainability, fairness, or green processing. Will this be good enough to help save the planet?
Monday, 25 October 2021, 17:15, Zoom Meeting
Please register at
https://mail-mm01.rrz.uni-hamburg.de/mailman/listinfo/kolloquium to receive the invitation data. The zoom access data will be mailed to attendees.