AI DJ System for Electronic Dance Music
Hao-Wei Huang, Muhammad Fadli, Achmad Kripton Nugraha, Chih-Wei Lin, Ray‐Guang Cheng · 2022
Disk jockey (DJ), the primary performer of electronic dance music, is often known to perform by observing the audience’s reaction and liven them by playing suitable EDM songs. DJ mixes two EDM songs without any breaks or silences, making the mix structurally coherent and seamless. However, DJs sometimes make mistakes during performances. An unsuitable song selection may occur due to the DJ’s inability to assess the audience’s reaction during a busy schedule. In the long run, hiring a DJ is quite expensive. A DJ is also limited in their ability to work continuously for long periods as a human being. Also, being a DJ requires expertise and experience, which most general music consumers lack. This paper proposes an automated DJ system featuring artificial intelligence (AI) named AI DJ system by combining action recognition, song selection, and beatmatching and equalizer mixing. The proposed system watches the audience’s reaction and selects the next song by using song similarity and choosing the similar key and suitable energy level. Additionally, beatmatching and equalizer mixing is applied to improve the song transition. Based on evaluations from 15 professional DJs in Taiwan using 400 EDM songs, the song mixes from the proposed system were confirmed to have good quality, with mean opinion score (MOS) as the metric.