Data & Matlab codes for "Performance Evaluation of Perceptible Impulsive Noise Detection Methods based on Auditory Models"
Paper submitted to EURASIP Journal on Audio, Speech, and Music Processing, 2024
Authors: Arda Ozdogru, Frantisek Rund, Karel Fliegel FEE CTU 2024
Can be found at https://mmtg.fel.cvut.cz/click-detection/ MATLAB Version: v9.14 (R2023a)
computer specifications: on Linux machines. The codes were run on a distributed computing infrastructure managed by MetaCentrum Virtual Organization https://metavo.metacentrum.cz/en/index.html
Some codes and materials are parts of JAES21: F. Rund, V. Vencovsk˝, and M. Semansk˝, An Evaluation of Click Detection Algorithms Against the Results of Listening Tests,î J. Audio Eng. Soc., vol. 69, no. 7/8, pp. 586-593, (2021 July.). https://doi.org/10.17743/jaes.2021.0020
Content of the folder (files and folders marked by ! are parts of other toolboxes and should be downloaded as specified):
--- experiment files ---
- bash_par/ contains the bash codes to run the related algorithms on Linux server in parallel (FOR PARALLEL COMPUTING).
- par_results/ results of the parallel computing. (.txt files are just for debugging)
- runX_lthY_BM_0.m files X = randomization number, Y = listening threshold, Z => 1 = Vencovsky's model, 0 = others (was necessary for timing)
- mX / mx_BM .m files codes called by the above codes.
- test_run.m to run after the training. needs to have par_results folder complete
- test_run.mat workspace after the test_run.m is completed. (final results are contained here) Ap error plot generation is in the second part of the code.
- other_analysis.m Analysis codes required for generated plots, ROC curve, logarithmic time comparison and ERBletHF parameter sweep results.
- best_param_extract_heatmap.py creates a heatmap of optimum parameters resulting in best A' scores over the experiments
--- common files ---
-
ScratchStimuliTest/ folder with the 90 wav files (44.1 kHz, 16 bit), 800 ms long, provided by gzmedia company. Taken from 3 vinyls, containing clicks due to manufacturing. - part of JAES21
-
subj_results.mat results of the listening tests as .mat file (Matlab) - part of JAES21
-
IR/ folder with the "impulse responses" of the used hearing models - response of the model to a click
-
! Aud_tbx/ M. Slaney Auditory Toolbox to download from https://engineering.purdue.edu/~malcolm/interval/1998-010/ (tested at ver 2) - necessary for Lyon's and Senef's models based algorithms
-
evaluation.m Matlab function for calculation of the comparison criteria (depreciated)
--- DRNL ---
- JAES_DRNL.m Matlab function for click detection using DRNL filter bank ear model based algorithm
- ! DRNL_MAP1_14 Matlab function for DRNL model, part of MAP toolbox, should be downloaded from https://github.com/rmeddis/MAP
- run_detect_DRNL.m Matlab script - run detection on the set of stimuli and compare it with the subjective results.
- Test_DRNL.m Matlab script - run run_detect_DRNL.m 10x to calculate average time elapsed for DRNL filter bank ear model based algorithm
--- Lyon ---
- JAES_lyon.m Matlab function for click detection using Lyon's model based algorithm (requires M. Slaney Auditory Toolbox in folder Aud_tbx)
- ! Aud_tbx/ M. Slaney Auditory Toolbox to download from https://engineering.purdue.edu/~malcolm/interval/1998-010/ (tested at ver 2) - necessary for Lyon's and Senef's models based algorithms
- run_detect_Lyon.m Matlab script - run detection on the set of stimuli and compare it with the subjective results.
- Test_Lyon.m Matlab script - run run_detect_Lyon.m 10x to calculate average time elapsed for Lyon's ear model based algorithm
--- Seneff ---
- JAES_seneff.m Matlab function for click detection using Seneff's model based algorithm (requires M. Slaney Auditory Toolbox in folder Aud_tbx)
- ! Aud_tbx/ M. Slaney Auditory Toolbox to download from https://engineering.purdue.edu/~malcolm/interval/1998-010/ (tested at ver 2) - necessary for Lyon's and Senef's models based algorithms
- run_detect_Seneff.m Matlab script - run detection on the set of stimuli and compare it with the subjective results.
- Test_Seneff.m Matlab script - run run_detect_Seneff.m 10x to calculate average time elapsed for Seneff's ear model based algorithm
--- Erblet ---
-
JAES_erblet.m Matlab function for click detection using Erblet based algorithm (requires Erblets functions and AMT installed)
-
! Erblets/ Folder with Matlab functions for Erblet transform to download from http://www.kfs.oeaw.ac.at/ICASSP2013_ERBlets - Need also AMT (https://www.amtoolbox.org/ , tested full 1.0.0 and 0.9.9 - ) - installed somewhere and then run amt_start.m
-
run_detect_erblet.m Matlab script - run detection on the set of stimuli and compare it with the subjective results.
-
Test_erblet.m Matlab script - run run_detect_erblet.m 10x to calculate average time elapsed for Erblet based algorithm
-
runAnalysisAR_JAES_f.m Matlab script - testing of the AR model based click detection method using the samples in * ScratchStimuliTest folder, including the performance (hits, false alarms, d') calculation - JAES21
-
fce_AR/ folder with functions for AR model based click detection method implementation JAES21
-
runAnalysismatch_JAES_f.m Matlab script - testing of the matched filter based click detection method using the samples in ScratchStimuliTest folder, including the performance (hits, false alarms, d') calculation JAES21
-
fce_matched/ folder with functions for matched filter based click detection method implementation JAES21
-
runAnalysisWavelet_JAES_f.m Matlab script - testing of the wavelet transform based click detection method using the samples in ScratchStimuliTest folder, including the performance (hits, false alarms, d') calculation JAES21
-
fce_wavelet/ folder with functions for wavelet transform based click detection method implementation and evaluation. JAES21
-
runBMmodelTEST_JAES_2.m Matlab script - testing of the Vencovsky's peripheral-ear model based click detection method using the samples in ScratchStimuliTest folder, including the performance (hits, false alarms, d') calculation JAES21
-
fce_audmodel/ folder with the functions for pheripheral-ear model click detection. JAES21