Stefan Goetze

South Westphalia University of Applied Sciences · Germany

Unclaimed · AI-structured profileORCID OpenAlex

Speech enhancement for hearing-impaired listeners: German applied-sciences researcher with a 2022–2026 output surge

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-02) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Not yet scored: the underlying output data is insufficient for this profile. The axis would measure publication volume, annual cadence, and authorship position; only per-year counts are quoted here, so no score or estimate is offered.

Impact—

Not yet scored: citation-impact data is insufficient for this profile. The axis would measure citation volume and scholarly influence; although an h-index of 22 and an i10-index of 44 are on file, no composite impact score is computed here.

Focus0/100

The focus axis scores 0 while the raw topics are tightly concentrated in one domain: Speech and Audio Processing (108) leads, with Speech Recognition and Synthesis (50), Advanced Adaptive Filtering Techniques (35), Hearing Loss and Rehabilitation (31), and Music and Audio Processing (31). The field distribution splits evenly — 0.5 speech processing, 0.5 computational hearing — across two adjacent labels, which may explain the zero score despite single-domain topical depth.

Collaboration44/100

Collaboration scores 44, a mid-range value. The recent window records 8 distinct co-authors against 28 works in the last three years, indicating small focused teams rather than broad networks. Recurring partners include Robert Sutherland, Jon Barker, Thomas Kuebert, Marko Lugger, Stefan Petrausch, Eline Borch Petersen, Juan Azcarreta Ortiz, and Buye Xu, and both reviewed papers are mid-position, non-corresponding contributions to consortium-scale efforts.

Activity98/100

Activity scores 98, the strongest axis in this profile, with 28 works across the last three years. The counts-by-year series shows a pronounced acceleration: 1 work in each of 2019 and 2020, then 9 in 2022, 19 in 2023, 15 in 2024, 9 in 2025, and 4 recorded for 2026 — a sustained high-cadence phase after the 2013–2016 plateau of 8–10 works per year.

Reproducibility55/100

Reproducibility scores 55, based on a small assessed base of 2 papers with a mean of 55. Both catalog-reviewed papers carry a medium reproducibility readiness rating: the CHiME-9 ECHI challenge paper relies on real-data subjective evaluation that is rigorous but hard to mirror exactly, and the multi-speaker extraction paper reports STOI gains from 0.55 to 0.60 with speech-only evaluation.

Research summary

Stefan Goetze (South Westphalia University of Applied Sciences, Germany; ORCID 0000-0003-1044-7343) works at the intersection of speech and audio signal processing and computational hearing. His OpenAlex record — name-resolved from public bibliographic data — lists an h-index of 22 and an i10-index of 44, with recorded output from 2006 onward and a marked recent acceleration: from one work in each of 2019 and 2020 to 9 in 2022, 19 in 2023, 15 in 2024, 9 in 2025, and 4 already logged in 2026, totaling 28 works in the last three years. Topic assignments confirm deep specialization: Speech and Audio Processing (108) leads, followed by Speech Recognition and Synthesis (50), Advanced Adaptive Filtering Techniques (35), Hearing Loss and Rehabilitation (31), and Music and Audio Processing (31). Recent Evolve-reviewed papers address hearing-impairment-driven enhancement of conversational speech: the CHiME-9 ECHI machine-learning challenge, whose real-data subjective evaluation is rigorous though objective-subjective mismatch remains under-analyzed, and multi-speaker extraction using a randomly VAD-masked loss that lifts STOI from 0.55 to 0.60. Both are mid-position, non-corresponding contributions with medium reproducibility readiness; the reproducibility mean across 2 assessed papers is 55. Frequent co-authors include Jon Barker, Robert Sutherland, Buye Xu, and other hearing-technology collaborators.

Collaboration pattern

Collaboration is real but comparatively compact: 8 distinct co-authors over the recent window against 28 works in three years, matching the mid-range collaboration score of 44. The recurring partner list — Robert Sutherland, Jon Barker, Thomas Kuebert, Marko Lugger, Stefan Petrausch, Eline Borch Petersen, Juan Azcarreta Ortiz, Buye Xu — reads as a stable speech-enhancement and hearing-technology circle, consistent with consortium-style challenge authorship; both reviewed papers are mid-position and non-corresponding.

Research keywords

speech enhancementtarget-speaker extractionmulti-speaker extractionhearing aidshearing loss and rehabilitationconversational speechspeech intelligibilityVAD-masked lossadaptive filteringspeech recognition and synthesismusic and audio processingCHiME-9 ECHI

Highlights

CHiME-9 ECHI: challenge design for hearing-impaired conversation enhancement
The 2026 challenge paper defines machine-learning enhancement of conversational speech for hearing-impaired listeners. Evolve's review credits rigorous real-data subjective evaluation, notes that winners are clear on hearing aids but contested on Aria, and flags that the objective-subjective mismatch is under-analyzed.
Randomly VAD-masked loss brings multi-speaker extraction to real conversations
The companion 2026 study proposes a randomly VAD-masked loss for multi-speaker extraction on real conversational data, lifting STOI from 0.55 to 0.60 in the CHiME-9 ECHI context. Gains are consistent, but absolute quality remains low and evaluation is speech-only.
Evaluation practice as the through-line: subjective, real-data, hearing-aid oriented
Across both reviewed papers the common thread is evaluation on real conversational data for hearing-impaired listeners — rigorous subjective listening tests in the challenge, and honest reporting of low absolute quality in the extraction study — with the objective-subjective mismatch named as the open problem.

Timeline

2006Earliest recorded output in the OpenAlex counts-by-year series: 4 works logged in 2006.
2015First-decade peak: 10 works in 2015, within a sustained 2013–2016 run of 8–10 works per year.
2019Trough of the series: only 1 recorded work in 2019 and again in 2020.
2023Series high: 19 recorded works in 2023, following 9 in 2022 and followed by 15 in 2024.
2026Two Evolve-reviewed papers appear: the CHiME-9 ECHI challenge on enhancing conversations for hearing impairment, and randomly VAD-masked multi-speaker extraction for real conversational speech; 4 works logged so far this year.

Papers in the Evolve catalog

Frequent co-authors

Robert Sutherland×2 Jon Barker×2 Thomas Kuebert×1 Marko Lugger×1 Stefan Petrausch×1 Eline Borch Petersen×1 Juan Azcarreta Ortiz×1 Buye Xu×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometric indicators are computed from OpenAlex only; the output and impact axes lack sufficient data and are left unscored.
Qualitative highlights rest on the two Evolve-reviewed catalog papers; the wider record is not individually assessed.
Identity is name-resolved from public bibliographic data (OpenAlex, ORCID 0000-0003-1044-7343); no biographical detail beyond affiliation and country is asserted.
Reproducibility draws on a small base: a mean of 55 over only 2 assessed papers, both rated medium.