Instructors:
Kristina Tesch
Displayed in timetable as:
MProj - SP
Language of instruction:
English
Min. | Max. participants:
- | 1
Comments/contents:
This master project provides an insight into current research topics in speech signal processing. The available problems include the separation of a speaker from a signal with multiple speakers, the localization of speakers in space, or the suppression of background noise. Example demos of comparable algorithms can be found at
https://www.youtube.com/watch?v=_wPZ2l12C-o&list=PLdsnmRiYvouBbSj_wxZhthEiwVp6TfDKN .
Students implement initial prototypes for the selected problem using machine learning methods (especially deep neural networks). Subsequently, we will aim at practical solutions (compliance with real-time conditions, implementation on portable devices). Finally, the strengths and weaknesses of the different approaches will be analyzed and evaluated on the basis of the implementations.
Learning objectives:
- expanding knowledge in the area of signal processing and machine learning
- independently comprehending a scientific topic
- translating scientific knowledge into practice within a project
- improving scientific writing and presentation skills
- improving teamwork
Didactic concept:
First, an introduction into the topic of (audio) signal processing will be given and relevant scientific literature will be discussed. Then, the concrete scope and the goals of the project will be finalized and the students will draft and present a work plan with milestones. Periodically, while working on the project, status updates will be presented and the work plan will be adopted accordingly when necessary.
Literature:
Literature recommendations will be provided at the start of the project.
