Recent years have seen some spectacular results in the field of sound and audio processing based on deep neural networks, including but not limited to common applications like speech recognition and generation, speaker recogntion, music analysis, noise filtering, and audio compression. In many cases, techniques developed in the context of computer vision can be applied or transferred to the audio domain. This success of deep learning approaches has attracted even more research activities, making it a highly vivid field.
This seminar will introduce basic ideas of processing audio data with deep neural networks and then focus on selected topics, like filtering, analysis and synthesis, in different domains, including speech, music, urban and animal sounds. The course will cover theoretical approaches as well as practical experiments with existing systems. Participants are assumed to have some basic experience in working with deep neural networks.
For COSMOS students not in Osnabrück: in case you would like to participate in this course, please contact us before semester start by email to check if we can find a way to make this work.