I am working on single channel speech enhancement. I have exploited statistical methods and deep learning models for this purpose. The source filter model for speech production can represent voiced frames in a compact and explicable manner. I improved statistical methods based on this model. Based on this decomposition, I investigated how to better preserve the speech harmonics by excitation signal manipulation and to enhance the speech intelligibility by an envelope estimate. Currently, I am considering how to extend these findings to deep learning methods. The multi-stage models are of great interest to me, where the complex task can be divided into smaller and easier parts and tackled individually.
I am a book worm. You can find pulp fictions, award-winning classics, as well as random nonfiction books in my readling list. I play story rich or creature collector video games a lot :D when I have vacations, I enjoy travelling and experiencing different things.
Since 2019: PhD student, Ghent University, Belgium 2016-2019: Master of Architectural and Civil Engineering, Tongji University, China 2012-2016: Bachelor of Civil Engineering, Tongji University, China