Metabuild Vocal Data Hackathon, 'Voice Separation Utilization Service' Wins Grand Prize

"Voice separation technology improves speech recognition by removing noise. It can be used in various ways in the entertainment industry such as music arrangement."

The "Rebellion from Below" team that submitted the idea of using voice separation technology won the grand prize at the "Artificial Intelligence Service and Business Model Idea Hackathon using multi-tonal vocal guide data" hosted by the Federation of Korean Music Performers, Metabuild, and Chilloen, and sponsored by the Ministry of Science and ICT and the National Information Society Agency. It is hosted by the Electronic Newspaper and ETEDU.

​

The winning team, consisting of students from KAIST Graduate School of AI, utilized a technology that separates audio signal data in which vocals and instruments are synthesized. For example, the data on singing in piano accompaniment are separated into piano and song audio. It uses meta data to create voice separation specialized for situations to enhance performance. It secures a sound separation source technology with good performance to remove noise or recycle it in various areas such as voice synthesis and vocal training.

​

The second prize was awarded to the "AI Logo Song Maker for Small Businesses Based on the Metaverse Platform" by the "What is this" team. The team, consisting of Yoon Tae-min, Sun Yu-jin, and Kim Sun-mi (general public), proposed an AI service that creates a logo song for small business owners using vocal data. It is available by selecting music genres, vocal tones, and store types through the metaverse's AI logo song maker "Logobot."

The "Vocal Helper" team consisting of students Park Hyun, Han Isaac and Park Min-je (Gwangju Institute of Science and Technology) was awarded with the participation award. The Vocal Helper team proposed an AI-based songwriting program. It uses various auxiliary functions using AI to support the production of individual music even if users do not have music expertise.

​

A total of 19 teams including university students participated in the hackathon, which was conducted as part of a data construction project for AI learning. After the first round of screening, nine teams participated in the second round of hackathon. They conducted online mentoring and hackathon online and non-face-to-face for two days from May 16-17. Kim Jae-eun, head of Saltlux, Moon Chae-hee, developer of Lotte Data Communication, and Lee Hae-seung, head of LuiTechnology Development Team, participated as mentors.​ 

Published: 21 Dec, 2021