{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,27]],"date-time":"2026-04-27T12:58:28Z","timestamp":1777294708903,"version":"3.51.4"},"reference-count":68,"publisher":"Acoustical Society of America (ASA)","issue":"4","content-domain":{"domain":["pubs.aip.org"],"crossmark-restriction":true},"short-container-title":[],"published-print":{"date-parts":[[2026,4,1]]},"abstract":"<jats:p>Methods for automatically assessing speech quality in real world environments are critical for robust human language technologies and assistive devices. Behavioral ratings (e.g., mean opinion scores) are considered the gold standard, but are susceptible to inter-rater variability, cannot be generalized, and are labor-intensive, thus limiting the acoustic challenges they can quantify. We present a scalable method for assessing speech quality: the self-supervised speech quality assessment model. First, we manipulated high-quality utterances, using acoustic challenges that emulate real-world degradation: filtering, reverberation, background noise, and digital compression. Second, we leveraged a pre-trained speech foundation model, WavLM, to derive cosine distances between the clean and degraded versions of each utterance in the embedding space. A transformer-based model was trained to predict these cosine distances, given only the degraded utterances. The trained model was evaluated on unseen corpora of synthetic mixtures, NISQA, and VOiCES. The self-supervised speech quality assessment model accurately predicts degradation cosine distances across a wide range of acoustic challenges and is aligned with behavioral ratings (mean opinion scores), automatic speech recognition performance and other important features (microphone distances). The model is available online.<\/jats:p>","DOI":"10.1121\/10.0043367","type":"journal-article","created":{"date-parts":[[2026,4,27]],"date-time":"2026-04-27T11:57:10Z","timestamp":1777291030000},"page":"3662-3673","update-policy":"https:\/\/doi.org\/10.1063\/aip-crossmark-policy-page","source":"Crossref","is-referenced-by-count":0,"title":["Self-supervised speech quality assessment (S3QA): Leveraging speech foundation models for a scalable speech quality metric"],"prefix":"10.1121","volume":"159","author":[{"ORCID":"https:\/\/orcid.org\/0009-0004-5583-0586","authenticated-orcid":false,"given":"Mattson","family":"Ogg","sequence":"first","affiliation":[{"name":"Johns Hopkins University Applied Physics Laboratory , Laurel, Maryland 20723,","place":["USA"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Caitlyn A.","family":"Bishop","sequence":"additional","affiliation":[{"name":"Johns Hopkins University Applied Physics Laboratory , Laurel, Maryland 20723,","place":["USA"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Han Gyol","family":"Yi","sequence":"additional","affiliation":[{"name":"Johns Hopkins University Applied Physics Laboratory , Laurel, Maryland 20723,","place":["USA"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sarah R.","family":"Robinson","sequence":"additional","affiliation":[{"name":"Johns Hopkins University Applied Physics Laboratory , Laurel, Maryland 20723,","place":["USA"]}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"231","published-online":{"date-parts":[[2026,4,27]]},"reference":[{"issue":"2","key":"2026042707570249800_c1","doi-asserted-by":"publisher","first-page":"114","DOI":"10.1109\/MSP.2014.2358871","article-title":"Objective quality and intelligibility prediction for users of assistive listening devices: Advantages and limitations of existing tools","volume":"32","year":"2015","journal-title":"IEEE Signal Process. Mag."},{"key":"2026042707570249800_c2","first-page":"749","article-title":"Perceptual Evaluation of Speech Quality (PESQ): A new method for speech quality assessment of telephone networks and codecs","year":"2001"},{"key":"2026042707570249800_c3","first-page":"4214","article-title":"A Short-Time Objective Intelligibility measure for time-frequency weighted noisy speech","year":"2010"},{"key":"2026042707570249800_c4","doi-asserted-by":"crossref","unstructured":"W.\n            Wang\n          , Y.Song, and S.Jha, \u201cGLOBE: A high-quality English corpus with global accents for zero-shot speaker adaptive text-to-speech,\u201d arXiv:2406.14875 (2024).","DOI":"10.21437\/Interspeech.2024-70"},{"key":"2026042707570249800_c5","first-page":"4218","article-title":"Common voice: A massively-multilingual speech corpus","author":"Calzolari","year":"2020"},{"key":"2026042707570249800_c6","doi-asserted-by":"crossref","unstructured":"H.\n            Zen\n          , V.Dang, R.Clark, Y.Zhang, R. J.Weiss, Y.Jia, and Z.Chen, \u201cLibriTTS: A corpus derived from LibriSpeech for text-to-speech,\u201d arXiv:1904.02882 (2019).","DOI":"10.21437\/Interspeech.2019-2441"},{"issue":"10","key":"2026042707570249800_c7","doi-asserted-by":"publisher","first-page":"1702","DOI":"10.1109\/TASLP.2018.2842159","article-title":"Supervised speech separation based on deep learning: An overview","volume":"26","year":"2018","journal-title":"IEEE\/ACM Trans. Audio. Speech Lang. Process."},{"key":"2026042707570249800_c8","doi-asserted-by":"publisher","first-page":"825","DOI":"10.1109\/TASLP.2020.2968738","article-title":"On loss functions for supervised monaural time-domain speech enhancement","volume":"28","year":"2020","journal-title":"IEEE\/ACM Trans. Audio. Speech Lang. Process."},{"key":"2026042707570249800_c9","first-page":"17022","article-title":"HiFi-GAN: Generative adversarial networks for efficient and high fidelity speech synthesis","volume-title":"Advances in Neural Information Processing Systems","year":"2020"},{"key":"2026042707570249800_c10","first-page":"1","article-title":"Freevc: Towards high-quality text-free one-shot voice conversion","year":"2023"},{"key":"2026042707570249800_c11","first-page":"3799","article-title":"{SMACK}: Semantically meaningful adversarial audio attack","year":"2023"},{"key":"2026042707570249800_c12","first-page":"166","article-title":"HiFi-GAN-2: Studio-quality speech enhancement via generative adversarial networks conditioned on acoustic features","year":"2021"},{"issue":"10","key":"2026042707570249800_c13","doi-asserted-by":"publisher","first-page":"2346","DOI":"10.1152\/jn.00372.2016","article-title":"Evidence of degraded representation of speech in noise, in the aging midbrain and cortex","volume":"116","year":"2016","journal-title":"J. Neurophysiol."},{"issue":"9","key":"2026042707570249800_c14","doi-asserted-by":"publisher","first-page":"3929","DOI":"10.1523\/JNEUROSCI.2908-14.2015","article-title":"Cortical activity predicts which older adults recognize speech in noise and when","volume":"35","year":"2015","journal-title":"J. Neurosci."},{"issue":"6","key":"2026042707570249800_c15","doi-asserted-by":"publisher","first-page":"1360","DOI":"10.1093\/cercor\/bhr211","article-title":"Word intelligibility and age predict visual cortex activity during word listening","volume":"22","year":"2012","journal-title":"Cereb. Cortex"},{"key":"2026042707570249800_c16","doi-asserted-by":"crossref","unstructured":"T.\n            Likhomanenko\n          , Q.Xu, V.Pratap, P.Tomasello, J.Kahn, G.Avidov, R.Collobert, and G.Synnaeve, \u201cRethinking evaluation in ASR: Are our models robust enough?,\u201d arXiv:2010.11745 (2021).","DOI":"10.21437\/Interspeech.2021-1758"},{"key":"2026042707570249800_c17","doi-asserted-by":"crossref","unstructured":"C.\n            Richey\n          , M. A.Barrios, Z.Armstrong, C.Bartels, H.Franco, M.Garciarena, A.Lawson, M. K.Nandwana, A.Stauffer, J. V.Hout, P.Gamble, J.Hetherly, C.Stephenson, and K.Ni, \u201cVoices obscured in complex environmental settings (VOICES) corpus,\u201d arXiv:1804.05053 (2018).","DOI":"10.21437\/Interspeech.2018-1454"},{"key":"2026042707570249800_c18","first-page":"818","article-title":"The Speakers in the Wild (SITW) Speaker Recognition Database","year":"2016"},{"key":"2026042707570249800_c19","first-page":"5695","article-title":"A comparative study of robustness of deep learning approaches for VAD","year":"2016"},{"key":"2026042707570249800_c20","doi-asserted-by":"publisher","first-page":"116","DOI":"10.1016\/j.neucom.2022.04.084","article-title":"A comprehensive empirical review of modern voice activity detection approaches for movies and TV shows","volume":"494","year":"2022","journal-title":"Neurocomputing"},{"key":"2026042707570249800_c21","article-title":"Brouhaha: Multi-task training for voice activity detection, speech-to-noise ratio, and C50 room acoustics estimation","year":"2023"},{"key":"2026042707570249800_c22","doi-asserted-by":"crossref","unstructured":"A.\n            Kumar\n          , K.Tan, Z.Ni, P.Manocha, X.Zhang, E.Henderson, and B.Xu, \u201cTorchAudio-Squim: Reference-less speech quality and intelligibility measures in TorchAudio,\u201d arXiv:2304.01448 (2023).","DOI":"10.1109\/ICASSP49357.2023.10096680"},{"key":"2026042707570249800_c23","first-page":"2142","article-title":"MetricNet: Towards improved modeling for non-intrusive speech quality assessment","year":"2021"},{"key":"2026042707570249800_c24","doi-asserted-by":"publisher","first-page":"54","DOI":"10.1109\/TASLP.2022.3205757","article-title":"Deep learning-based non-intrusive multi-objective speech assessment model with cross-domain features","volume":"31","year":"2023","journal-title":"IEEE\/ACM Trans. Audio. Speech Lang. Process."},{"issue":"2","key":"2026042707570249800_c25","doi-asserted-by":"publisher","first-page":"213","DOI":"10.1007\/s00530-014-0446-1","article-title":"Mean opinion score (MOS) revisited: Methods and applications, limitations and alternatives","volume":"22","year":"2016","journal-title":"Multimedia Syst."},{"key":"2026042707570249800_c26","first-page":"6493","article-title":"DNSMOS: A non-intrusive perceptual objective speech quality metric to evaluate noise suppressors","year":"2021"},{"key":"2026042707570249800_c27","first-page":"2127","article-title":"NISQA: A deep CNN-self-attention model for multidimensional speech quality prediction with crowdsourced datasets","year":"2021"},{"key":"2026042707570249800_c28","doi-asserted-by":"crossref","unstructured":"P.\n            Manocha\n           and A.Kumar, \u201cSpeech quality assessment through MOS using non-matching references,\u201d arXiv:2206.12285 (2022).","DOI":"10.21437\/Interspeech.2022-407"},{"key":"2026042707570249800_c29","first-page":"8442","article-title":"Generalization ability of MOS prediction networks","year":"2022"},{"key":"2026042707570249800_c30","doi-asserted-by":"crossref","unstructured":"S.\n            Chen\n          , C.Wang, Z.Chen, Y.Wu, S.Liu, Z.Chen, J.Li, N.Kanda, T.Yoshioka, X.Xiao, J.Wu, L.Zhou, S.Ren, Y.Qian, Y.Qian, J.Wu, M.Zeng, X.Yu, and F.Wei, \u201cWavLM: Large-scale self-supervised pre-training for full stack speech processing,\u201d arXiv:2110.13900 (2022).","DOI":"10.1109\/JSTSP.2022.3188113"},{"key":"2026042707570249800_c31","doi-asserted-by":"crossref","unstructured":"A.\n            Pasad\n          , C.-M.Chien, S.Settle, and K.Livescu, \u201cWhat do self-supervised speech models know about words?,\u201d arXiv:2307.00162 (2024).","DOI":"10.1162\/tacl_a_00656"},{"key":"2026042707570249800_c32","first-page":"1","article-title":"Improved decoding of attentional selection in multi-talker environments with self-supervised learned speech representation","year":"2023"},{"key":"2026042707570249800_c33","doi-asserted-by":"publisher","first-page":"21895","DOI":"10.48550\/arXiv.2305.11863","article-title":"Scaling laws for language encoding models in fMRI","volume":"36","year":"2023","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"2026042707570249800_c34","first-page":"2225","article-title":"Do self-supervised speech and language models extract similar representations as human brain?","year":"2024"},{"key":"2026042707570249800_c35","first-page":"3849","article-title":"No-reference speech intelligibility prediction leveraging a noisy-speech ASR pre-trained model","year":"2024"},{"key":"2026042707570249800_c36","unstructured":"H.\n            Wang\n          , J.Zhao, Y.Yang, S.Liu, J.Chen, Y.Zhang, S.Zhao, J.Li, J.Zhou, H.Sun, Y.Lu, and Y.Qin, \u201cSpeechLLM-as-judges: Towards general and interpretable speech quality evaluation,\u201d arXiv:2510.14664 (2025)."},{"key":"2026042707570249800_c37","unstructured":"M.\n            Monjur\n           and S.Nirjon, \u201cSpeechQualityLLM: LLM-based multimodal assessment of speech quality,\u201d arXiv:2512.08238 (2025)."},{"key":"2026042707570249800_c38","first-page":"591","article-title":"Blind estimation of the speech transmission index for speech quality prediction","year":"2018"},{"key":"2026042707570249800_c39","unstructured":"J.\n            Yamagishi\n          , C.Veaux, and K.MacDonald, \u201cCSTR VCTK corpus: English multi-speaker corpus for CSTR Voice Cloning Toolkit (version 0.92),\u201d in IEEE International Conference on Acoustics, Speech and Signal Processing (2019), http:\/\/web.ku.edu\/\u223cidea\/readings\/rainbow.htm."},{"key":"2026042707570249800_c40","first-page":"2756","article-title":"AISHELL-3: A Multi-speaker Mandarin TTS corpus","year":"2021"},{"key":"2026042707570249800_c41","unstructured":"L.\n            Zheng\n          , F.Hong, and W.Xu, \u201cA text-dependent speaker verification application framework based on Chinese numerical string corpus,\u201d arXiv:2312.01645 (2023)."},{"key":"2026042707570249800_c42","doi-asserted-by":"crossref","first-page":"188","DOI":"10.1007\/978-3-031-40498-6","article-title":"CML-TTS: A multilingual dataset for speech synthesis in low-resource languages","volume-title":"Text, Speech, and Dialogue","author":"Ek\u0161tein","year":"2023"},{"key":"2026042707570249800_c43","unstructured":"M.\n            Cartwright\n          , J.Cramer, A. E. M.Mendez, Y.Wang, H.-H.Wu, V.Lostanlen, M.Fuentes, G.Dove, C.Mydlarz, J.Salamon, O.Nov, and J. P.Bello, \u201cSONYC-UST-V2: An urban sound tagging dataset with spatiotemporal context,\u201d arXiv:2009.05188 (2020)."},{"key":"2026042707570249800_c44","first-page":"1","article-title":"Environmental sound classification with convolutional neural networks","year":"2015"},{"key":"2026042707570249800_c45","first-page":"1041","article-title":"A dataset and taxonomy for urban sound research","year":"2014"},{"key":"2026042707570249800_c46","unstructured":"D.\n            Stowell\n           and M. D.Plumbley, \u201cAn open dataset for research on audio field recording archives: freefield1010,\u201d arXiv:1309.5275 (2013)."},{"key":"2026042707570249800_c47","unstructured":"E.\n            Fonseca\n          , M.Plakal, F.Font, D. P. W.Ellis, X.Favory, J.Pons, and X.Serra, \u201cGeneral-purpose tagging of freesound audio with audioset labels: Task description, dataset, and baseline,\u201d arXiv:1807.09902 (2018)."},{"key":"2026042707570249800_c48","doi-asserted-by":"crossref","unstructured":"E.\n            Fonseca\n          , M.Plakal, F.Font, D. P. W.Ellis, and X.Serra, \u201cAudio tagging with noisy labels and minimal supervision,\u201d arXiv:1906.02975 (2020).","DOI":"10.33682\/w13e-5v06"},{"key":"2026042707570249800_c49","doi-asserted-by":"publisher","first-page":"829","DOI":"10.1109\/TASLP.2021.3133208","article-title":"FSD50K: An open dataset of human-labeled sound events","volume":"30","year":"2022","journal-title":"IEEE\/ACM Trans. Audio. Speech Lang. Process."},{"key":"2026042707570249800_c50","doi-asserted-by":"crossref","unstructured":"G.\n            Wichern\n          , J.Antonioni, M.Flynn, L. R.Zhu, E.McQuinn, D.Crow, E.Manilow, and J.Le Roux, \u201cWHAM!: Extending speech separation to noisy environments,\u201d arXiv:1907.01160 (2019).","DOI":"10.21437\/Interspeech.2019-2821"},{"key":"2026042707570249800_c51","first-page":"776","article-title":"Audio set: An ontology and human-labeled dataset for audio events","year":"2017"},{"key":"2026042707570249800_c52","first-page":"5220","article-title":"A study on data augmentation of reverberant speech for robust speech recognition","year":"2017"},{"key":"2026042707570249800_c53","first-page":"165","article-title":"Database of omnidirectional and B-format room impulse responses","year":"2010"},{"issue":"4","key":"2026042707570249800_c54","doi-asserted-by":"publisher","first-page":"863","DOI":"10.1109\/JSTSP.2019.2917582","article-title":"Building and evaluation of a real room impulse response dataset","volume":"13","year":"2019","journal-title":"IEEE J. Sel. Top. Signal Process."},{"issue":"48","key":"2026042707570249800_c55","first-page":"E7856","article-title":"Statistics of natural reverberation enable perceptual separation of sound and space","volume":"113","year":"2016","journal-title":"Proc. Natl. Acad. Sci. U.S.A."},{"key":"2026042707570249800_c56","first-page":"18","article-title":"librosa: Audio and music signal analysis in Python","year":"2015"},{"key":"2026042707570249800_c57","article-title":"pyloudnorm: A simple yet flexible loudness meter in Python","year":"2021"},{"key":"2026042707570249800_c58","first-page":"6982","article-title":"Torchaudio: Building blocks for audio and speech processing","year":"2022"},{"key":"2026042707570249800_c59","doi-asserted-by":"publisher","first-page":"3451","DOI":"10.1109\/TASLP.2021.3122291","article-title":"HuBERT: Self-supervised speech representation learning by masked prediction of hidden units","volume":"29","year":"2021","journal-title":"IEEE\/ACM Trans. Audio Speech Lang. Process."},{"key":"2026042707570249800_c60","first-page":"28492","article-title":"Robust speech recognition via large-scale weak supervision","year":"2023"},{"key":"2026042707570249800_c61","doi-asserted-by":"crossref","unstructured":"S.\n            Zhang\n          , J.Yi, Z.Tian, J.Tao, Y. T.Yeung, and L.Deng, \u201cReducing language context confusion for end-to-end code-switching automatic speech recognition,\u201d arXiv:2201.12155 (2022).","DOI":"10.21437\/Interspeech.2022-10286"},{"issue":"70","key":"2026042707570249800_c62","doi-asserted-by":"publisher","first-page":"4101","DOI":"10.21105\/joss.04101","article-title":"TorchMetrics: Measuring reproducibility in PyTorch","volume":"7","year":"2022","journal-title":"J. Open Source Software"},{"key":"2026042707570249800_c63","first-page":"1","article-title":"Self-supervised transformer model training for a sleep-EEG foundation model","year":"2024"},{"key":"2026042707570249800_c64","first-page":"914","article-title":"Layer-wise analysis of a self-supervised speech representation model","year":"2021"},{"key":"2026042707570249800_c65","first-page":"1","article-title":"Comparative layer-wise analysis of self-supervised speech models","year":"2023"},{"key":"2026042707570249800_c66","unstructured":"M.\n            Ogg\n          , \u201cSelf-supervised convolutional audio models are flexible acoustic feature learners: A domain specificity and transfer-learning study,\u201d arXiv:2502.02366 (2025)."},{"key":"2026042707570249800_c67","doi-asserted-by":"publisher","first-page":"137","DOI":"10.1109\/TASLP.2022.3221007","article-title":"BYOL for audio: Exploring pre-trained general-purpose audio representations","volume":"31","year":"2023","journal-title":"IEEE\/ACM Trans. Audio. Speech Lang. Process."},{"key":"2026042707570249800_c68","first-page":"5206","article-title":"Librispeech: An ASR corpus based on public domain audio books","year":"2015"}],"container-title":["The Journal of the Acoustical Society of America"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/pubs.aip.org\/asa\/jasa\/article-pdf\/159\/4\/3662\/20985518\/3662_1_10.0043367.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/pubs.aip.org\/asa\/jasa\/article-pdf\/159\/4\/3662\/20985518\/3662_1_10.0043367.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,27]],"date-time":"2026-04-27T11:57:32Z","timestamp":1777291052000},"score":1,"resource":{"primary":{"URL":"https:\/\/pubs.aip.org\/jasa\/article\/159\/4\/3662\/3387984\/Self-supervised-speech-quality-assessment-S3QA"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,1]]},"references-count":68,"journal-issue":{"issue":"4","published-print":{"date-parts":[[2026,4,1]]}},"URL":"https:\/\/doi.org\/10.1121\/10.0043367","relation":{},"ISSN":["1520-8524"],"issn-type":[{"value":"1520-8524","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026,4]]},"published":{"date-parts":[[2026,4,1]]}}}