@audio/shift-hpss

import hpss from '@audio/shift-hpss'

Fitzgerald median-filter harmonic/percussive separation. Time-axis and frequency-axis medians produce soft Wiener masks splitting the spectrogram. Harmonic component is vocoder-shifted; percussive component passes through with original phase.

hpss(audio, { ratio: 1.5 })
hpss(audio, { ratio: 1.5, hpssTimeWidth: 31, hpssFreqWidth: 31 })

Preserves percussive onset locations (unshifted) and harmonic pitch (shifted).
Destroys signal quality at ambiguous mask boundaries (leakage in both directions).

Best overall shift score — keeping percussion unshifted sidesteps most artifacts. It also has the best attack correlation in the whole collection (0.998): the percussive component's phase is never touched, so a plucked-string attack survives more faithfully than even the time-domain similarity-search methods. Alias 0.052 is residual harmonic energy leaking through the percussive mask.

Use when: Mixed music where drums should stay stationary while melody shifts.
Not for: Solo tonal material (unnecessary separation overhead).

Stream

let write = hpss({ ratio: 1.5 })
let out = write(inputBlock)
let tail = write()  // flush

hpss buffers the whole input; write(chunk) accumulates and the batch algorithm runs once on write() (flush, no argument), returning everything at once.

Data

Input is a Float32Array (mono) or an array of Float32Array channels ([left, right, ...]) — anything else throws TypeError. ratio also accepts a function t => ratio (seconds from stream start) or a Float32Array breakpoint envelope (resampled across the input via ratioDuration, default the input's own duration) for time-varying pitch.

Part of @audio/shift — the shift family umbrella. This README is generated from the umbrella docs.

MIT © audiojs