Music Scene Description

This chapter introduces a research approach called ‘music scene description’ [232], [225], [228], where the goal is to build a computer system that can understand musical audio signals at the level of untrained human listeners without trying to extract every musical note from music. People listening to music can easily hum the melody, clap hands in time to the musical beat, notice a phrase being repeated, and find chorus sections. The brain mechanisms underlying these abilities, however, are not yet well understood. In addition, it has been difficult to implement these abilities on a computer system, although a system with them is useful in various applications such as music information retrieval, music production/editing, and music interfaces. It is therefore an important challenge to build a music scene description system that can understand complex real-world music signals like those recorded on commercially distributed compact discs (CDs).