@audio/auditory-erb

ERB-scale band centers and bandwidths (Glasberg & Moore 1990) — no filtering

npm install @audio/auditory-erb

import erbBank from '@audio/auditory-erb'

Band centers equally spaced on the ERB-number scale E(f) = 21.4*log10(4.37*f/1000 + 1), inverted back to Hz, each paired with its equivalent rectangular bandwidth ERB(f) = 24.7*(4.37*f/1000 + 1) (Glasberg & Moore, "Derivation of auditory filter shapes from notched-noise data", Hearing Research 47, 1990).

Design function, not a stream processor: erbBank(fs, opts) → bands. Positional fs first — a deliberate divergence from the filter/eq family's fn(data, params) in-place convention. No buffer in, no buffer out — and unlike its Bark/mel/octave siblings, it does no filtering at all: it returns {fc, erb} pairs only, no coefs. Pair with @audio/auditory-gammatone (one call per band, using each fc) to actually filter audio.

erbBank(44100)                    // ERB-spaced bands, defaults
erbBank(44100, { density: 2 })    // double band density

| Param | Default | | |---|---|---| | fmin | 50 | Lowest band center Hz | | fmax | min(16000, fs/2) | Highest band center Hz | | density | 1 | Bands per ERB — step along the ERB-number axis is 1/density |

Use when: perceptually-uniform frequency sampling for auditory-model front-ends or excitation-pattern analysis, without committing to a filter shape.

Part of @audio/auditory — the auditory family umbrella.

MIT © audiojs