THE AUDITORY MODELING TOOLBOX

Applies to version: 1.3.0

View the code

Go to function

KING2019 - Modulation filterbank (based on nonlinear processing)

Usage

outsig = king2019(insig,fs,basef)

Input parameters

insig input acoustic signal.
fs sampling rate (Hz).
basef base frequency of the analysis (Hz).

Output parameters

outsig output signal
fc center frequencies filterbank
mfc center frequencies modulation filterbank
step struct containing intermediate model outputs

Description

king2019(insig,fs,basef) computes the internal representation around the frequency basef of the signal insig sampled with a frequency of fs Hz. outsig is a matrix of size [length fc mfc].

[outsig,fc,mfc,step]=king2019(...) additionally returns the centre frequencies of the filter bank and the center frequencies of the modulation filterbank, and 'steps' is a structure containing the intermediate model outputs.

The model consists of the following stages:

  1. a gammatone filter bank with 1-erb spaced filtes.
  2. a compression (exponential or broken-stick) stage applied to each individual gammatone filter. In these stages is extremely relevant to use the correct calibration level (default is dboffset = 100), given that the keyval 'compression_knee_dB' is referenced to the dboffset. Note that if other dboffset values are used (e.g., dboffset=94 dB), the knee point will in fact start compressing at higher levels.
  3. an envelope extraction stage done by half-wave rectification followed by low-pass filtering to 1500 Hz.
  4. an simplified adaptation stage simulated as a first-order high-pass filter
  5. a modulation filterbank

Many parameters (keyvals) and flags can be optionally changed and/or switched

References:

A. King, L. Varnet, and C. Lorenzi. Accounting for masking of frequency modulation by amplitude modulation with the modulation filter-bank concept. J. Acoust. Soc. Am., 145(2277), 2019.