A Music Exploration Interface Based on Vocal Timbre and Pitch in Popular Music

This paper proposes an interface that enables music exploration focusing on two factors of singing voices, vocal timbre and pitch, which are useful in finding singing voices that match users' preferences. The proposed interface uses a two-dimensional map to visualize songs being explored and locate them according to timbre or pitch similarities of their singing voices. Since similar songs are located closely on the map, users can visually find a singing voice similar to their favorite singing voices. In addition to the location, the interface uses the color of each song on the map to visualize an additional factor related to characteristics of singing voices, such as acoustic features or words describing singing voices (e.g., "Clear"). Prior to developing the interface, we conducted a questionnaire survey with 20 subjects and confirmed that both vocal timbre and pitch are important when listening to music. The proposed interface was implemented with 102 songs, and a user study was conducted with 60 subjects.