THE AUDITORY MODELING TOOLBOX

Applies to version: 1.6.0

View the code

Go to function

EURICH2022
Binaural detection model based on two-hemisphere code and interaural incoherence interference

Usage:

[processed, mpar] = eurich2022(mRef, mTest,mpar)

Input parameters:

mRef Binaural reference signal
mTest Binaural test signal
mpar Model parameters

Output parameters:

processed Processed model outputs (reference and test, concatenated)
mpar Modified model parameters

Description:

eurich2022 is a binaural detection model based on two-hemisphere code. It involves interaural incoherence interference. For binaural reference and test signals, it computes the complex correlation coefficient (gamma) representing the two-hemisphere encoding found in mammals' medial superior olives (MSO). To account for degraded binaural detection for maskers with frequency-dependent interaural coherence, an across-frequency interference mechanisms of interaural incoherence is introduced. Additionally, the model computes the DC power to account for energetic unmasking. eurich2022_processing computes these features based on a binaural signal. Then, eurich2022_decision derives the model's sensitivity based on these computed features.

mpar is an optional structure with the following parameters:

  • fs: sampling frequency of model
  • Filters_per_ERB_aud: spacing of peripheral filter central frequencies in ERB
  • GT_bandwidth_factor: factor of gammatone filter bandwidth relative to "standard" (i.e. ERB = 79 Hz @500Hz)
  • GT_lowest_center_frequency: lower bound of gammatone filterbank
  • GT_highest_center_frequency: higher bound of gammatone filterbank
  • GT_fix_center_frequency: fixed center frequency of one of the gammatone filterbank's filters
  • GT_filterorder: filter order of gammatone filterbank
  • interference_sigma: std of the Gaussian window used for the across-frequency incoherence interference
  • iKernelThresh: threshold above which a value of the Gaussian filter window is used
  • rho_max upper limit of encoded interaural coherence (i.e. internal noise)
  • monaural_internal_noise_sigma: std of the Gaussian, level-dependent internal noise of the monaural feature
  • binaural_internal_noise_sigma: std of the Gaussian internal noise of the binaural feature
  • target_channel: index of the channel to be selected for processing in eurich2022_decision

References:

B. Eurich, J. Encke, S. D. Ewert, and M. Dietz. Lower interaural coherence in off-signal bands impairs binaural detection. The Journal of the Acoustical Society of America, 151(6):3927--3936, 06 2022. [ DOI ]