THE AUDITORY MODELING TOOLBOX

Applies to version: 1.6.0

View the code

Go to function

BAUMGARTNER2021
Sound externalization based on multiple static cues

Usage:

[E,cues,cueLabels] = baumgartner2021( target,template )

Input parameters:

target Binaural impulse responses referring to the directional transfer function(s) (DFTs) of the target sound(s). Option 1: given in SOFA format -> sagittal plane DTFs will be extracted internally. Option 2: binaural impulse responses of all available listener-specific DTFs of the sagittal plane with the size: (time x direction x channel).
template Binaural impulse responses of all available listener-specific DTFs of the sagittal plane referring to the perceived lateral angle of the target sound. Format as target.

Output parameters:

E predicted degree of externalization
cues outcomes of individual cues
cueLabels cue labels; cell array with 1st col. denoting acronyms and 2nd col. for descriptions

Description:

baumgartner2021(...) is a model framework for auditory externalization perception. It enables to probe the contribution of cue-specific expectation errors and to contrast dynamic versus static strategies for combining those errors within static listening environments.

baumgartner2021 accepts the following optional parameters:

'cueWeights',cW

Set the weights of individual cues to determine the final externalization score. Cue-specific weights (entered as a vector) are ordered as follows:

1 monaural spectral similarity (MSS)

2 interaural spectral similarity of ILDs (ISS)

3 spectral standard deviation of monaural gradients (MSSD)

4 spectral standard deviation of ILDs (ISSD)

5 interaural broadband time-intensity coherence (ITIT)

6 interaural coherence (IC)

7 monaural intensity difference (MI)

8 temporal standard deviation of ILDs (ITSD).

Default weights are 0.6 for MSS, 0.4 for ISS, and 0 for all others.

'S',S Set the cue-specific sensitivity parameter to S. 1/S represents the slope of sigmoidal mapping function. Vector order equivalent to cueWeights. Default values are determined by the weighted average sensitivities determined in Baumgartner and Majdak (2020) - run exp_baumgartner2021('tab2') to show them.
'lat',lat Set the apparent lateral angle of the target sound to lat. Default value is 0 degree (median SP).
'range',c1 Set the range factor of the externalization scores to c1. Default value is 3.78 from Hassager et al. (2016).
'offset',c2 Set the offset of the externalization score to c2. Default value is 1 from Hassager et al. (2016).
'ILD_JND',L Set the just noticeable ILD difference to L from the internal template. Default value is 1 (dB).
'ITD_JND',T Set the just noticeable ITD difference to T from the internal template. Default value is 20e-6 (s).

baumgartner2021 accepts the following flags:

'LTA' Looser-takes-all strategy: Model selects minimal predicted externalization scores across cues with weights larger than zero.
'MTA' Median-takes-all strategy: Model selects median predicted externalization scores across cues with weights larger than zero.
'WTA' Winner-takes-all strategy: Model selects maximal predicted externalization scores across cues with weights larger than zero.

Requirements:

  1. SOFA API from http://sourceforge.net/projects/sofacoustics for Matlab (in e.g. thirdparty/SOFA)
  1. Circular Statistics Toolbox from http://www.mathworks.com/matlabcentral/fileexchange/10676-circular-statistics-toolbox--directional-statistics-

References:

R. Baumgartner and P. Majdak. Decision making in auditory externalization perception: model predictions for static conditions. Acta Acustica, 5:59, 2021. Publisher: EDP Sciences. [ DOI ]