eurich2024
Efficient monaural and binaural model for audio quality assessment (eMoBi-Q)

Usage:

[features, mpar] = eurich2024(mRef, mTest, fs);
[features, mpar] = eurich2024(mRef, mTest, fs, mpar)

Input parameters:

`mRef`	Binaural audio reference signal. Size: (time x ear).
`mTest`	Binaural audio test signal. Size: (time x ear).
`fs`	Sampling rate (in Hz) of mRef and mTest.
`mpar`	Optional structure with the model parameters. Defaults: as in arg_eurich2024. The following fields are supported: fs: Sampling frequency (in Hz) of model. GT_Filters_per_ERB_aud: Spacing (in ERB) of peripheral filter central frequencies. GT_bandwidth_factor: Factor of the Gammatone filter bandwidth (relative to the bandwidth of 79 Hz at center frequency of 500 Hz). GT_lowest_center_frequency: Center frequency (in Hz) of lowest filter in the Gammatone filterbank. GT_highest_center_frequency: Center frequency (in Hz) of highest filter in the Gammatone filterbank. GT_fix_center_frequency: Center frequency (in Hz) of the fixed filter in the Gammatone filterbank. GT_filterorder: Filter order of the Gammatone filterbank. env_lowpass_fc: Cutoff frequency (in Hz) of the envelope lowpass filter. env_lowpass_n: Order of the envelope lowpass filter. mso_rolloff: Frequency threshold (in Hz) below that Gammas are calculated based on fine structure and above that Gammas are calculated based on envelopes. interference_sigma: Standard deviation 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 (represents existence of internal noise). ild_lim: Limit for the ILD (in dB). FrameLen: Length (in samples) of consecutive rectangular time frames.

Output parameters:

features

Structure containing the features calculated by the model, containing:

d_gamma: Gamma features. Size: (m x n), with m being the time frames and n the frequency bands.
d_ild: ILD features. Size: (m x n).
d_snr: SNR features. Size: (m x n).

mpar

Structure with the updated model parameters.

Description:

Note: if mpar is not provided, it is loaded from arg_eurich2024.

References:

B. Eurich, S. D. Ewert, M. Dietz, and T. Biberger. A computationally efficient model for combined assessment of monaural and binaural audio quality. Journal of the Audio Engineering Society, 100(4):2381--2396, June 2024. [ DOI ]