2026-10-07T01:51:16Z https://api.figshare.com/v2/oai

oai:figshare.com:article/32415357 2026-09-25T09:15:49Z category_30388 category_30319 category_28861 category_28846 portal_379 item_type_3 month_year_09_2026

Dataset for: "A model-based analysis of modulation masking effects on vocoded speech intelligibility" Cathrina Veigel (17096374) Helia Relaño-Iborra (18180632) Andrew J. Oxenham (18544993) Torsten Dau (6213869) Sensory processes, perception and performance Testing, assessment and psychometrics Speech recognition Modelling and simulation Hearing research Speech intelligibility modelling vocoded speech intelligibility modulation masking This dataset supports the findings reported in:Veigel, C., Relaño-Iborra H., Oxenham A.J., Dau, T.; A model-based analysis of modulation masking effects on vocoded speech intelligibility. J. Acoust. Soc. Am. September 2026; 160 (3): 2090–2104. <a href="https://doi.org/10.1121/10.0046394" target="_blank" rel="noreferrer">https://doi.org/10.1121/10.0046394</a> The README gives a detailed description of the files, structure and variables of the dataset. The behavioral data replicated and extended the findings from Oxenham and Kreft (2014).The dataset contains behavioral data of 14 native English speaking participants with normal-hearing. The data collected was:<ul><li>Population data: age, gender, English dialect</li><li>Pure-tone audiograms</li><li>Percent correctly understood words for half-AzBio sentence lists (Spahr et al, 2012) in noise, both unprocessed and tone-vocoded.</li></ul>Furthermore the dataset contains all masker stimuli, the vocoder MATLAB functions (same as in Oxenham and Kreft, 2014) as well as the model predictions of the data, generated with the speech-based envelope power spectrum model with a correlation back-end (sEPSMcorr, Relaño-Iborra et al, 2016, modified in Veigel et al, 2026). The model including its versions are available <a href="https://bitbucket.org/heliaib/sepsm-corr/downloads/" target="_blank" rel="noreferrer">here</a>.The dataset is licensed under <a href="https://creativecommons.org/licenses/by/4.0/" target="_blank" rel="noreferrer">Creative Commons Attribution 4.0 (CC BY 4.0)</a><h3 dir="ltr">Files Descriptions</h3>population_data_audiograms.csv: age, gender, dialect and audiograms of the test subjects.<ul><li>subject: subject number. only subjects after data selection are included (see *data_collection_and_selection.pdf*)</li><li>gender: gender that the subject identifed as (no options were given) m = masculin f = feminin</li><li>age: age of the subjects at the moment the data was collected (in years)</li><li>dialect: self-reported English dialect of the subject</li><li>R_0.25kHz: pure tone threshold at 0.25 kHz in dB HL for the right ear</li><li>R_0.5kHz: pure tone threshold at 0.5 kHz in dB HL for the right ear</li><li>R_1kHz: pure tone threshold at 1 kHz in dB HL for the right ear</li><li>R_2kHz: pure tone threshold at 2 kHz in dB HL for the right ear</li><li>R_4kHz: pure tone threshold at 4 kHz in dB HL for the right ear</li><li>R_8kHz: pure tone threshold at 8 kHz in dB HL for the right ear</li><li>L_0.25kHz: pure tone threshold at 0.25 kHz in dB HL for the left ear</li><li>L_0.5kHz: pure tone threshold at 0.5 kHz in dB HL for the left ear</li><li>L_1kHz: pure tone threshold at 1 kHz in dB HL for the left ear</li><li>L_2kHz: pure tone threshold at 2 kHz in dB HL for the left ear</li><li>L_4kHz: pure tone threshold at 4 kHz in dB HL for the left ear</li><li>L_8kHz: pure tone threshold at 8 kHz in dB HL for the left ear</li><li>PTA4(dB HL): pure tone average threshold of the frequencies 0.5,1,2,4 kHz of left and right ear</li></ul>Behavioral_Data_Pcorrect_avg_per_list.csv: Speech intelligibility was measured as percent correctly understood words in different processing and masker conditions. Percent correct scores were averaged per half-AzBio lists (10 sentences).<ul><li>subject: subject number. Includes the subjects after data selection (see *data_collection_and_selection.pdf*)</li><li>condition: processing condition. unpro = unprocessed, bp = bandpass-filtered (cutoffs of the vocoder), voc1 = 16-channel tone-vocoded without current spread, or voc2 = 16-channel tone-vocoded with simulated current spread</li><li>masker: masker type. SSN = speech shaped noise, SSNbp = bandpass-filtered speech shaped noise (cutoffs of the vocoder, measured only in unpro condition), PT = pure-tone complex, MT = modulated pure-tone complex, or PT+2LF = pure-tone complex with two additional low frequencies (measured only in unpro condition)</li><li>Pcorrect: averaged word correct percentage (in %) for the measured list (=10 sentences) A detailed description of the masker types and the processing conditions are available in the article Veigel et al (2026).</li></ul>data_collection_and_selection.pdf: explanation of which data was included in the final, published dataset, as some of the data had to be excluded due to technical and time issues.masker_files.zip: wav-files of the 5 masker types used in Veigel et al (2026).<ul><li>The files are 20 s long, sampled with a sampling frequency of 22050 Hz. Assume that all maskers have an rms-level of -20 dB, eventhough the maskers SSNbp, PT and MT actually have lower rms-levels, as they were created to fit the SSN energy per vocoder channel. Energy below and above the vocoder passband are not represented in those maskers - thus the maskers have same rms per vocoder channel, but the overall rms is different and should not be normalized in order to replicate the study. This leads to relative overall SNR differences across maskers in the unprocessed condition, but not in the other conditions (for more information see Veigel et al, 2026). </li></ul>vocoder_code.zip: matlab functions to create the same vocoded signals as in this study. Vocoder implementations are the same as used in Oxenham and Kreft (2014). Published with permission.<ul><li>demod_vocoder.m: tone-excited envelope vocoder. Takes the signal to vocode (x), the vocoder channel center frequencies (channel_cfs, in this study: [333 455 540 642 762 906 1076 1278 1518 1803 2142 2544 3022 3590 4264 6665] Hz), the vocoder lowpass-filter cutoff (env_lpco, here: 50 Hz) and the sampling frequency (fs) as input and returns the vocoded singal (x_final_output).</li><li>demod_vocoder_spread.m: tone-excited envelope vocoder that simulates current spread across vocoder channels. Takes x, channel_cfs, env_lpco and fs (same as in no-spread vocoder), as well as the the amount of attenuation of the spectral smearing across vocoder channels in dB/oct (spread, in this study: 8) as input and returns the vocoded singal (x_final_output).</li><li>bandpass_filter.m: bandpass-filter design. Takes sampling frequency fs and filter order as inputs and returns filter coefficients a and b</li><li>folder functions: help functions for the vocoder implementations</li></ul>Predicted_Pcorrect_avg_over_100sent.csv: sEPSMcorr (Relaño-Iborra et al, 2016, modified in Veigel et al, 2026) predictions. Find descriptions of the maskers types and processing conditions above and in Veigel et al (2026)<ul><li>condition: processing condition. unpro, bp, voc1 or voc2</li><li>masker: masker type. SSN, PT or MT</li><li>pred-Pcorrect: predictions of sEPSMcorr3, word correct percentage (in %), averaged across 100 sentences</li></ul><h3 dir="ltr">Ethical Statement</h3>All participants received financial compensation for their time and provided written informed consent. The study was approved by the Science-Ethics Committee for the Capital Region of Denmark (reference: H-16036391).<h3 dir="ltr">References</h3>Spahr, A. J., Dorman, M. F., Litvak, L. M., Van Wie, S., Gifford, R. H., Loizou, P. C., Loiselle, L. M., Oakes, T., Cook,S.: Development and Validation of the AzBio Sentence Lists. Ear and Hearing 33(1):p 112-117, January 2012. |DOI: <a href="https://doi.org/10.1097/AUD.0b013e31822c2549" target="_blank" rel="noreferrer">10.1097/AUD.0b013e31822c2549</a>Relaño-Iborra, H., May, T., Zaar, J., Scheidiger, C., Dau, T.: Predicting speech intelligibility based on a correlation metric in the envelope power spectrum domain. J. Acoust. Soc. Am. 1 October 2016; 140 (4): 2670–2679.<a href="https://doi.org/10.1121/1.4964505" target="_blank" rel="noreferrer">https://doi.org/10.1121/1.4964505</a>Steinmetzger, K., Zaar, J., Relaño-Iborra, H., Rosen, S., Dau, T.: Predicting the effects of periodicity on theintelligibility of masked speech: An evaluation of different modelling approaches and their limitations. J.Acoust. Soc. Am. 1 October 2019; 146 (4): 2562–2576. <a href="https://doi.org/10.1121/1.5129050" target="_blank" rel="noreferrer">https://doi.org/10.1121/1.5129050</a>Oxenham A. J., Kreft H.A.: Speech Perception in Tones and Noise via Cochlear Implants Reveals Influence of Spectral Resolution on Temporal Processing. Trends in Hearing. 2014;18. doi:<a href="https://doi.org/10.1177/2331216514553783" target="_blank" rel="noreferrer">10.1177/2331216514553783</a><h3 dir="ltr">Citation and links</h3><h4 dir="ltr">Cite this dataset:</h4>Veigel, C., Relaño-Iborra, H., Oxenham, A., Dau, T.: Dataset for "A model-based analysis of modulation masking effects on vocoded speech intelligibility". Technical University of Denmark. Dataset | DOI: <a href="https://doi.org/10.11583/DTU.32415357" target="_blank" rel="noreferrer">https://doi.org/10.11583/DTU.32415357</a><h4 dir="ltr">Corresponding article:</h4>Veigel, C., Relaño-Iborra H., Oxenham A.J., Dau, T.; A model-based analysis of modulation masking effects on vocoded speech intelligibility. J. Acoust. Soc. Am. September 2026; 160 (3): 2090–2104. <a href="https://doi.org/10.1121/10.0046394" target="_blank" rel="noreferrer">https://doi.org/10.1121/10.0046394</a><h4 dir="ltr">Corresponding preprint:</h4>Veigel, C., Relaño-Iborra, H., Oxenham, A., Dau, T.: A model-based analysis of modulation masking effects on vocoded speech intelligibility. PsyArXiv Preprint. March 27, 2026 | DOI:<a href="https://doi.org/10.31234/osf.io/3h9pv_v1" target="_blank" rel="noreferrer">https://doi.org/10.31234/osf.io/3h9pv_v1</a><h3 dir="ltr">Acknowledgments and funding</h3>This work has received funding by the Center for Applied Hearing Research (CAHR) and the William Demant Foundation (Grant Number 23-4855). Andrew J. Oxenham's contributions were supported by NIH grant R01 DC023255. 2026-09-25T09:15:49Z Dataset Dataset 10.11583/DTU.32415357.v2 https://figshare.com/articles/dataset/Dataset_for_A_model-based_analysis_of_modulation_masking_effects_on_vocoded_speech_intelligibility_/32415357 CC BY 4.0