University of Sheffield researcher behind CHiME-9 ECHI: speech enhancement for hearing-impaired conversation
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
0
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—
Unscored — the indexed window contains only two works (countsByYear 2026: 2; recent3yWorks 2), too thin to measure total publication volume, venue mix or output trajectory. This axis would aggregate career-wide papers across years and venues.
Impact—
Unscored — citation indicators are absent in this snapshot (h-index 0, i10-index 0) and both papers are brand-new 2026 preprints, so influence cannot yet be measured. This axis would track citations, uptake in the hearing-aid enhancement community, and downstream challenge participation.
Focus0/100
The radar records focus at 0/100, the lowest axis in this profile. Taken literally it sits oddly against the raw field distribution, which is actually narrow: speech processing and speech and audio signal processing (computational hearing) split 0.5/0.5, and the four OpenAlex topics (counts 2, 2, 1, 1) all cluster around conversational speech enhancement for hearing impairment. The 0 most plausibly reflects the very small indexed sample (two works) rather than genuine topic scatter.
Collaboration44/100
Collaboration scores 44/100, consistent with a moderate-size challenge consortium pattern. The snapshot records 8 distinct coauthors, and the frequent-coauthor list spans Stefan Goetze, Jon Barker, Thomas Kuebert, Marko Lugger, Stefan Petrausch, Eline Borch Petersen, Juan Azcarreta Ortiz and Buye Xu. Both catalog papers are first-authored but community challenge outputs, so coauthor counts reflect the CHiME-9 ECHI organizing scale more than long-running bilateral ties.
Activity32/100
Activity scores 32/100 on a thin recent record: recent3yWorks is 2 and countsByYear shows a single year, 2026, with 2 outputs. Both are arXiv preprints rather than venue publications, and no earlier indexed years appear in the snapshot. The low score therefore marks the start of an indexed publishing record rather than a documented slowdown — the window contains too little history to judge cadence.
Reproducibility55/100
Reproducibility scores 55/100 (reproMean 55 across reproPapers 2), the strongest signal in this early profile. Both catalog papers carry Evolve's medium repro-readiness rating. The ECHI challenge paper is credited with rigorous real-data subjective evaluation, while the multi-speaker extraction study reports a concrete STOI gain (0.55 to 0.60) from a randomly VAD-masked loss; reviewer caveats include under-analyzed objective–subjective mismatch and speech-only evaluation.
Research summary
Robert Sutherland is a speech-processing researcher at the University of Sheffield (GB) whose indexed record concentrates on machine learning for conversational speech enhancement aimed at hearing-impaired listeners. The Evolve catalog holds two 2026 arXiv preprints, both first-authored: the CHiME-9 ECHI challenge paper, which defines a machine-learning challenge for enhancing conversations to address hearing impairment, and a methods study of multi-speaker extraction on real conversational audio. OpenAlex topic assignment spans speech recognition and synthesis (2), emotion and mood recognition (2), hearing loss and rehabilitation (1) and speech and audio processing (1), with a field distribution split evenly (0.5/0.5) between speech processing and speech and audio signal processing (computational hearing). The recorded network covers 8 distinct coauthors, including Stefan Goetze, Jon Barker, Thomas Kuebert, Marko Lugger, Stefan Petrausch, Eline Borch Petersen, Juan Azcarreta Ortiz and Buye Xu, a scale consistent with consortium-style challenge authorship. Citation indicators (h-index, i10-index) currently read 0 in the snapshot, leaving output and impact unscored; reproducibility averages 55 across the two reviewed papers, and Evolve's verdicts converge on evaluation methodology — rigorous subjective protocols, but objective–subjective mismatch under-analyzed and evaluation still speech-only.
Collaboration pattern
Collaboration follows a challenge-consortium pattern: 8 distinct coauthors are recorded (radar collaboration 44/100), and the frequent list — Stefan Goetze, Jon Barker, Thomas Kuebert, Marko Lugger, Stefan Petrausch, Eline Borch Petersen, Juan Azcarreta Ortiz and Buye Xu — maps onto the CHiME-9 ECHI community around both 2026 first-authored papers. The profile shows consortium-scale, multi-party authorship rather than a small bilateral cluster; no long-term dyadic pattern can yet be inferred from two papers.
Research keywords
Highlights
Designing CHiME-9 ECHI: a community challenge for hearing-grade conversation enhancement
Randomly VAD-masked loss pushes real-conversation extraction forward
Reviews converge on evaluation methodology as the open question
Across both first-authored 2026 papers, Evolve's verdicts converge on evaluation: the ECHI subjective protocol is rigorous but the objective–subjective mismatch is under-analyzed, and the multi-speaker extraction evaluation is speech-only — a consistent methodological frontier for this line of work.
Timeline
- 2026First-author paper launching the CHiME-9 ECHI machine-learning challenge for enhancing conversations to address hearing impairment appears on arXiv.
- 2026First-author study on the challenges of multi-speaker extraction for real conversational speech enhancement is posted to arXiv, introducing a randomly VAD-masked loss with STOI rising from 0.55 to 0.60.
- 2026OpenAlex records both outputs under a single year (countsByYear 2026: 2) with 8 distinct coauthors, and Evolve reviews rate both papers medium for reproducibility readiness.
Papers in the Evolve catalog
Frequent co-authors
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.
- Bibliometrics come from an OpenAlex snapshot only; h-index and i10-index read 0 and the output and impact radar axes are unscored.
- The profile rests on two 2026 arXiv preprints, both first-authored; no peer-reviewed venue record or ORCID linkage is available.
- Reproducibility (55/100) averages Evolve reviews of just two papers, both rated medium repro-readiness; treat it as provisional.
- "Robert Sutherland" is a common name; this profile is name-resolved from public bibliographic data (OpenAlex) and should be confirmed against ORCID before reuse.