Concurrent Estimation of Chords and Keys from Audio

This paper proposes a new method for local key and chord estimation from audio signals. A harmonic content of the musical piece is first extracted by computing a set of chroma vectors. Correlation with fixed chord and key templates then selects a set of key/chord pairs for every frame. A weighted acyclic harmonic graph is then built with these pairs as vertices, and the use of a musical distance to weigh its edges. Finally, the output sequences of chords and keys are obtained by finding the best path in the graph. The proposed system allows a mutual and beneficial chord and key estimation. It is evaluated on a corpus composed of Beatles songs for both the local key estimation and chord recognition tasks. Results show that it performs better than state-of-the art chord analysis algorithms while providing a more complete harmonic analysis.