Date：2025/5/2 15:30-17:00
Location: R103, CSIE
Speakers: Prof. Wee Kheng Leow
Host: 洪一平教授
Abstract:
Style-specific harmonization aims at generating chords that match a given melody and specific style. Existing methods for generating chords have some shortcomings. Most of them are trained on a large training set of a specific genre such as pop or jazz. To accommodate different styles, they need to train on different style-specific training sets, increasing the total training time. Moreover, existing methods cannot cater to cadence and pitch fluctuation. Many recent deep neural networks (DNNs) for harmonization are tested only on 8-bar phrases instead of complete music pieces. Therefore, they cannot handle global music structure. This talk presents a new style-specific harmonization method called FlexChord that overcomes the shortcomings of existing methods. FlexChord achieves style specificity by producing chords that match the style of selected reference music. It can perform chord generation with as little as a single music piece, and it does not require a large reference set. It applies direct optimization during chord generation that does not require separate training. Thus, it can easily produce chords of a specific style. In addition, FlexChord can handle cadence, pitch fluctuation and the global structure of full-length melodies. Test results show that FlexChord's harmonization is better than those of comparable DNNs and is close to that of human harmonization.
Biography:
Wee Kheng Leow obtained his B.Sc. and M.Sc. in National University of Singapore in 1985 and 1989, respectively. After obtaining Ph.D. in Computer Science at The University of Texas at Austin in 1994, he joined the Department of Computer Science, National University of Singapore as an Associate Professor. He has worked on areas including neural networks, image processing, computer vision and 3D modelling. His current research focuses on medical image analysis and surgery simulation and planning. He has published many technical papers in journals and conferences. He has also been awarded three U.S. patents and a Singapore patent. He has collaborated widely with many local and overseas institutions, including National Taiwan University and National Taiwan University Hospital. His current active collaborators include Singapore General Hospital, National University Hospital, Peking University First Hospital and a Singapore subsidiary of Furui Medical Science Company at Beijing.