楊奕軒 Yi-Hsuan Yang

Professor, Department of Electrical Engineering, National Taiwan University

Major Research Areas:

Music information research; Artificial intelligence; Machine learning; Music generation

Research Summary:

We work on projects that develop machine learning and deep learning models to address research problems that are about music signals, including music analysis (e.g., automatic music transcription, source separation, and classification) and automatic music generation (including the generation of symbolic MIDI as well as music audio).  Over the recent years, we have the pleasure to propose various AI models that represent the state-of-the-art in automatic music generation at the time, such as the "Pop Music Transformer" and "Compound Word Transformer" for expressive pop piano music generation, and the "KaraSinger" for melody-free singing voice generation.  We hope that our research cares more about the quality of the generated music, rather than simply publishing papers.  We enjoy the creative process that harmonizes knowledge in the two fairly different fields: music and machine learning.  

Prior to joining my alma mater NTU EE as a Full Professor, I spent 11.5 years as an Assistant/Associate Research Fellow at the Research Center for Information Technology Innovation at Academia Sinica (https://twmusicai.github.io/); and 4 years as a full-time Chief Music Scientist at the Taiwan AI Labs (https://ailabs.tw/). 

You can find more information of our research from the materials below.  I am also active on Twitter (https://twitter.com/affige_yang).

• https://affige.github.io/publication.html
• https://case.ntu.edu.tw/sciactivity/20211019/
• https://case.ntu.edu.tw/blog/?p=38833
• https://www.slideshare.net/affige/automatic-music-composition-with-transformers-jan-2021
• https://alumni.ee.ntu.edu.tw/?p=5853