THE AUDITORY MODELING TOOLBOX

Applies to version: 1.6.0

View the code

Go to function

bramslow2004
Loudness model considering hearing loss (AUDMOD)

Usage:

[audout, fc] = bramslow2004(insig,fs);
[audout, fc] = bramslow2004(insig,fs,...);

Input parameters:

insig Acoustic signal (in Pa).
fs Sampling rate (in Hz).

Output parameters:

audout

Structure containing the following:

  • EC: Row vector with ERB rates (in Cams) of the filters (typically 3 to 32). Size: NoChan (defined by the input parameter 'NoChan', default: 30 channels).
  • fc: Row vector with center frequencies (in Hz) of the filters. Size: NoChan.
  • ERB_R: Row vector with the excitation pattern at the hearing thresholds (in dB). Size: NoChan.
  • ERB_UCL: Row vector with the excitation pattern at the uncomfortabe levels (UCLs, in dB). Size: NoChan.
  • FFT_Power: Column vector with total powers (in dB) of individual signal frames. Size: frames, being floor(insig/In_FrmSize) defined by the input parameter 'In_FrmSize' (default: 8192 samples).
  • ERB_SPL: Matrix with SPLs in ERB bands, used for level dependencies of filterbank (in dB). Size: (frames x NoChan).
  • ERB_Power: Vector with the total power (in dB) in ERB bands. Size: (frames x 1).
  • Roex_SPL: Matrix with SPLs (in dB) of the excitation patterns (from the Roex filters). Size: (frames x NoChan).
  • Total_Excitation: Column vector with the total SPL (in dB) of the excitation of the individual signal frames. Size: frames.
  • ERB_Loudness: Matrix with the specific loudness (in sone) per ERB band and the individual signal frames. Size: (frames x NoChan).
  • Loudness: Column vector with the total loudness (in sone) of the individual signal frames. Size: frames.
fc Output channel center frequencies (in Hz). The same as audout.fc.

Description:

audout = bramslow2004(insig,fs) computes the internal representation of the signal insig sampled at a frequency of fs. The main output is audout.Loudness which returns the loudness (in sone) of the individual frames of the input signal insig.

bramslow2004(..) also accepts the following flags among many others:

'no_debug' No display of debugging information. Default.
'debug' Display additional debugging information. Opposite of 'no_debug'.

See also other flags and key-value pairs in arg_bramslow2004.

[audout,fc] = bramslow2004(...) additionally returns the center frequencies of the filterbank.

The processing blocks in the model are as follows:

  • FFT spectrum, e.g., 512 frequencies to obtain a power spectrum.
  • Corrections for sound-field type, coupler type, and static transmission factors (e.g., middle ear), applied in the frequency domain.
  • The signal power is determined in 1 ERB wide bands, by summing the power spectrum (f) within the limits of each band. These powers are used to adjust the filterbank.
  • The FFT power spectrum is multiplied by a filterbank, consisting of 30 auditory filters whose shape depends on hearing loss and on the signal power. The filterbank concept is based on work from Moore, Glasberg, Patterson and others at the University of Cambridge and uses rounded exponential (roex) filters. The roex filterbank output is called the excitation pattern.
  • The parameters for hearing loss (thresholds) are converted from dB HL to dB SPL and used to modify frequency selectivity in the filterbank and sensitivity in the loudness function.
  • The roex filterbank output (E) is passed on to the specific loudness function converting excitation in each channel to specific loudness (N'). The total loudness of an incoming signal is calculated by summing the specific loudness across bands.

References:

L. Bramsløw Nielsen. An Auditory Model with Hearing Loss. Technical report, Eriksholm Research Centre, Snekkersten, 1993. [ .pdf ]

L. Bramsløw. An objective estimate of the perceived quality of reproduced sound in normal and impaired hearing. Acta Acustica united with Acustica, 90(6):1007--1018, 2004. [ http ]

L. Bramsløw. An auditory loudness model with hearing loss. In Baltic-Nordic Acoustics Meeting, pages 318--323, 2024. [ .html ]

B. R. Glasberg and B. Moore. Derivation of auditory filter shapes from notched-noise data. Hearing Research, 47(1-2):103--138, 1990.

Build with Bootstrap