THE AUDITORY MODELING TOOLBOX

Applies to version: 1.5.0

View the code

Go to function

LLADO2022 - Neural network localization

Usage:

[y_est] = llado2022(ir);
[y_est] = llado2022(ir, stim);
[y_est] = llado2022(ir, stim, fs);
[y_est] = llado2022(ir, stim, fs, NN_pretrained);

Input parameters:

ir Impulse response according to the following matrix dimensions [direction x time x channel/ear]

Output parameters:

y_est Estimated values for perceived direction and position uncertainty.

Description:

LLADO2022(...) is a model for estimating the effect of head-worn devices on frontal horizontal localisation. A neural network (NN) was trained using binaural features of a dummy head wearing different head-worn devices to predict the data from a perceptual test using the same devices. If you want to use your own data, please find in the script 'demo_llado2022' the whole procedure.

Optional input parameters:

'stim' stimulus. If empty, 250 ms of pink noise
'fs' (DEFAULT = 48000)
'NN_pretrained' if empty, a pretrained NN is used.

To see details or to train a new NN, please see the script demo_llado2022

References:

Lladó, Pedro, Hyvärinen, Petteri, and Pulkki, Ville. Auditory model-based estimation of the effect of head-worn devices on frontal horizontal localisation. Acta Acust., 6:1, 2022. [ DOI | http ]