@audio/noise-estimate

@audio/noise-estimate

Noise power-spectral-density estimation — the noise floor that spectral denoisers subtract.

Three stateful estimators over STFT magnitude frames. Feed them per-frame; read .psd (a Float64Array of N/2+1 bins) whenever you need the current noise estimate.

import { minStats } from '@audio/noise-estimate'
import { stftAnalyse } from '@audio/stft'

let est = minStats(1024, { D: 96 })          // half = frameSize/2
stftAnalyse(signal, mag => est.update(mag), { frameSize: 2048 })
let noisePsd = est.psd                        // drive Wiener / MMSE / OM-LSA gain

minStats(half, opts?)

Minimum Statistics (Martin 2001) — tracks a rolling D-frame minimum of the smoothed PSD per bin, scaled by a bias factor so the minimum estimates E{|N|²} rather than the lower tail. opts: D (window frames, default 96 ≈ 1.5 s), alpha (PSD smoothing, 0.7), bias (1.5). Returns { psd, update(mag) }.

imcra(half, opts?)

Improved Minima-Controlled Recursive Averaging (Cohen 2003) — updates the noise PSD only where speech is absent, gated by a speech-presence probability (supply your own via update(mag, spp), or it derives one from the smoothed-to-minimum ratio). opts: alpha (0.92), alphaD (0.85), beta (1.47). Returns { psd, update(mag, sppOverride?) }.

noiseProfile(data, opts?)

One-shot baseline: averages |X|² over a quiet segment (opts.from/opts.to samples). Returns a Float64Array PSD. Use when you can point at a known noise-only region.

Notes

STFT via @audio/stft; pairs with @audio/vad's spp(). Also re-exported from @audio/denoise. MIT.