I am a Research Scientist II at Adobe Research. I received a PhD in Computer Science and a master in Music at University of California San Diego (UCSD). Previously, I interned at Adobe, Mitsubishi Electric Research Laboratories (MERL), Apple and other companies focusing on the music and general audio research.
My main interest lies in Audio Representation Learning and Generation, and its application to downstreaming tasks. It is an inter-disciplinary among music, general audio, and technology.
Several projects I lead during my PhD and current workflows:
- Audio Representation Learning
- Language-Audio Modeling
- Audio Understanding
- Music Recommendation
- Audio Source Separation
- Universal Audio Source Separation
- Musical Instrument Source Separation
- Chorale Music Source Separation
- Music Generation & Music Information Retrieval
- Controllability in Music Generation
- Chord Recognition
- Singing Melody Extraction
Highlighted Projects with regard to the above research I lead or serve as a main contributor:
- FLAM: Frame-wise Language-Audio Modeling
- CLAP: Contrastive Language-Audio Pretraining
- HTS-AT: Hierarchical Audio Transformer
- PromptSep: Generative Audio Separation via Multimodal Prompting
- Zero-Shot Audio Source Separation
- BACHI: Chord Recognition on Symbolic Music
- POP909-CL: POP909 with human annotations
- POP909: A Dataset for Pop Music Arrangement
- Music SketchNet: Controllable Algorithmic Music Composition
- TONet: A Singing Melody Extraction Framework
- Choral Music Separation
I was the website maintainer of New Interfaces for Musical Expression. My alias "Knut" comes from one of my favorite composers Knut Nystedt.
