THE AUDITORY MODELING TOOLBOX

Applies to version: 1.4.0

View the code

Go to function

BARUMERLI2023_FEATUREEXTRACTION - extract binaural and monaural cues from SOFA object

Usage:

[template, target] = barumerli2023_featureextraction(sofa_obj)

Input parameters:

sofa_obj Struct in SOFA format with DTFs

Output parameters:

template internal templates with specific feature points
target (optional) preprocessed target struct

Description:

barumerli2023_featureextraction(...) computes temporally integrated spectral magnitude profiles, itd and ild.

Additional input parameters:

'fs' Sampling rate. Default: 48kHz.
'flow' Low frequency for auditory bands. Default: 700Hz
'fhigh' High frequency for auditory bands. Default: 18kHz
'space' Auditory bands spacing. Default: 1 ERB
'targ_az' column vector to select the binaural stimulus in the templates with the specified azimuth in degree shall be used as target. Use in combination with targ_el. Default: [].
'targ_el' column vector to select the binaural stimulus in the templates with the specified elevation in degree shall be used as target. Use in combination with targ_az. Default: [].
'source_ir' Specify a custom sound source to be convolved with HRTFs. The default consider broadband noise. Default: []
'source_fs' Sampling rate of the source. Default: 0 Hz

Further, cache flags (see amt_cache) and plot flags can be specified:

'template' Compute only templates.
'target' Compute only targets.
'pge' Use spectral gradients as monaural cues.
'dtf' Use spectral amplitudes as monaural cues.
'monaural_none' Do not use monaural cues.
'reijniers' Compute feature space as in Reijniers et al. 2014
'source' Use sound source provided with the parameter source_ir to compute targets.