A research team from Human Language Technology Lab was ranked first in the following tasks at the Music Information Retrieval Evaluation eXchange (MIREX) International Benchmarking Competition 2020.
- Automatic Lyrics-to-Audio Alignment
- Automatic Lyrics Transcription
The winning team comprised of Ph.D. candidate Ms. Xiaoxue Gao and Research Fellow Dr. Chitralekha Gupta supervised by Professor Li Haizhou from NUS Electrical and Computer Engineering.
Automatic lyrics-to-audio alignment seeks to time synchronize between music audio and its written lyrics, that produces synchronized on-screen lyrics display for karaoke singing. The task of lyrics transcription seeks to convert sung words in music audio to text just like speech recognition for speech. The ECE team has won top position in both tasks in 2020 and has been ranked first for automatic lyrics-to-audio alignment task two years in a row. The results were announced at the ISMIR 2020 conference on 16 October 2020. Established in 2005, MIREX has fostered advancements both in specific areas of music information retrieval and in the general understanding of how music information retrieval systems and algorithms are to be evaluated.
The winning Lyrics-to-Audio Alignment system NUS Auto Lyrix Align is now available online as an interactive web interface: https://autolyrixalign.hltnus.org/. This technology is useful for music TV production, music streaming services, music labels, and karaoke services.
Video Demonstration of the NUS Auto Lyrix Align Technology:
To find out more on the research activities of Human Language Technology Lab, please visit: http://ece.nus.edu.sg/hlt/
