Chroma Interval Content as a Key-Independent Harmonic Progression Feature

Abstract

This paper introduces a novel chroma-based harmonic feature called Chroma Interval Content (CIC), which extends Directional Interval Content (DIC) vectors to audio data. This feature represents key-independent harmonic progressions, but unlike the Dynamic Chroma feature vector it represents pitch-class energy motions based on a symbolic voice-leading approach, and can be computed more efficiently (in time $\mathcal{O}(N\log N)$ as opposed to $\mathcal{O}(N^{2}))$. We present theoretical properties of Chroma Interval Content vectors and explore the expressive power of CIC both in representing isolated chord progressions, establishing links to its symbolic counterpart DIC, as well as in specific harmony-related MIR tasks, such as key-independent search for chord progressions and classification of music datasets according to harmonic diversity.