@audio/mir-chroma

Chroma (pitch-class profile) — a 12-D vector where each bin holds the energy attributed to one pitch class (C, C♯, …, B), folding all octaves together.

npm install @audio/mir-chroma

import chroma from '@audio/mir-chroma'

let c = chroma(data, { fs: 44100 })   // Float64Array[12], L1-normalized (sums to 1)

Two methods:

• 'pcp' (default) — classical Fujishima (1999). Each spectral bin is mapped to its nearest pitch class by log-frequency rounding and power-accumulated.
• 'nnls' — Mauch & Dixon (2010) NNLS Chroma. Fits the observed spectrum as a nonnegative combination of synthetic pitch-tone profiles (fundamental + 1/h-decaying overtones) via multiplicative NMF updates, then folds pitch activations into pitch classes. Cleaner on polyphonic audio — suppresses octave and harmonic confusion that PCP is prone to.

Options: - fs — sample rate (default 44100, Hz) · method — 'pcp' or 'nnls' (default 'pcp') · minFreq/maxFreq — analysis band for 'pcp' (default 65/2093, Hz, ≈C2–C7) · harmonics — partials per pitch template for 'nnls' (default 8) · iterations — NMF update steps for 'nnls' (default 30)

data must be a power-of-2-length window (throws otherwise) — chroma operates on one FFT frame per call; slide it yourself for a chromagram.

Also exported as an audio.js stat manifest (./audio) — chromagram frames over time plus their normalized mean.

Part of @audio/mir.