Rasmus Kongsgaard Olsson
Victoria University of Wellington · New Zealand
Speech enhancement researcher spanning blind source separation, generative models, and embedded NPU deployment
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
7
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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: this snapshot lacks the total-publication and output-trajectory aggregates the axis needs. It would measure career-wide publishing volume, venue mix, and output rate over time, complementing the raw year counts already on file.
Impact—
Not yet scored: per-paper citation totals and field-normalized comparisons are absent from the snapshot, so influence cannot be rated. The axis would weigh raw citations and normalized impact, with the h-index of 7 and i10-index of 6 as the only raw signals on file.
Focus0/100
Scores 0/100 because the OpenAlex field split is perfectly even: 0.5 for speech signal processing (machine learning) and 0.5 for speech signal processing / embedded machine learning. In practice the two labels are adjacent, and topic counts stay coherent — speech and audio processing (14), blind source separation (10), speech recognition and synthesis (9), with sparse and compressive sensing (3) and image enhancement (3) as the lightest areas.
Collaboration36/100
Scores 36/100 on a compact network: only 5 distinct co-authors appear in the recent window, and five names recur — Liang Xu, Diego Caviedes-Nozal, W. Bastiaan Kleijn, Longfei Felix Yan, and Clément Laroche. Both reviewed 2026 papers place Olsson in the last author slot without corresponding-author status, a pattern consistent with senior involvement in student-led projects rather than broad multi-team networking.
Activity47/100
Scores 47/100. The snapshot records 4 works in the last three years, with the two reviewed 2026 arXiv papers among that year's 4 works, up from a single work in 2022; earlier clusters show 3–4 works per year across 2004, 2006, and 2007. The pattern reads as a reactivated research line concentrated in speech enhancement rather than continuous high-volume output.
Reproducibility55/100
Scores 55/100, the mean of Evolve's readiness grades across 2 reviewed papers, both rated medium. The grounding metrics are concrete — PESQ 3.01 at 4.83 ms/frame on an MCU for the LiSenNet redesign, and WER-first state-of-the-art claims for DriftSE — but both verdicts flag single-run means without error bars, seeds, or uncertainty, which caps the score.
Research summary
Rasmus Kongsgaard Olsson is a speech signal processing researcher at Victoria University of Wellington, New Zealand. This profile is name-resolved from public OpenAlex bibliographic data; no ORCID is on file. The snapshot records an h-index of 7 and an i10-index of 6, with year-by-year counts of 3 works in 2004, 4 in 2006, 4 in 2007, a single work in 2022, and 4 in 2026. Topically the record concentrates on speech and audio processing (14 works), blind source separation techniques (10), and speech recognition and synthesis (9), with smaller footprints in sparse and compressive sensing (3) and image enhancement (3). Recent field allocation splits evenly, 0.5 and 0.5, between machine-learning speech signal processing and embedded machine learning for speech. The last three years show 4 works with 5 distinct co-authors. Evolve has reviewed two 2026 arXiv papers listing Olsson in the last author position: a redesign of LiSenNet into an int8 NPU model reaching PESQ 3.01 at 4.83 ms/frame on a microcontroller, and DriftSE, a one-step latent-drifting generative speech-enhancement model with WER-first state-of-the-art results. Both carry medium repro-readiness; the verdicts credit credible engineering while flagging single-run means without error bars or seeds. Recurring co-authors include Liang Xu, Diego Caviedes-Nozal, W. Bastiaan Kleijn, Longfei Felix Yan, and Clément Laroche.
Collaboration pattern
Collaboration is compact: 5 distinct co-authors appear in the recent window, and five names recur across the record — Liang Xu, Diego Caviedes-Nozal, W. Bastiaan Kleijn, Longfei Felix Yan, and Clément Laroche — yielding a collaboration score of 36/100. On both reviewed 2026 papers Olsson takes the last-listed author position without corresponding-author status, a pattern consistent with senior or supervisory involvement in student-led work rather than large multi-institution consortia.
Research keywords
Highlights
LiSenNet rebuilt for microcontrollers: int8 NPU model at PESQ 3.01 and 4.83 ms/frame
DriftSE: one-step latent-drifting generative speech enhancement
Senior-author anchor across two 2026 speech-enhancement tracks
Olsson holds the last author position on both Evolve-reviewed 2026 papers without corresponding-author status, one targeting embedded deployment and the other generative modeling. The pairing mirrors the snapshot's even 0.5/0.5 field split between machine-learning speech processing and embedded machine learning for speech.
Timeline
- 2004OpenAlex records 3 works this year, the earliest cluster in the snapshot.
- 20064 works indexed, tying 2007 as the snapshot's busiest early year.
- 20074 works indexed, closing out the early 2004–2007 run before the record gaps.
- 2022A single work indexed — a sparse bridge year in the snapshot.
- 20264 works recorded, including the two Evolve-reviewed arXiv papers — the LiSenNet redesign for embedded speech enhancement and the DriftSE generative enhancer — both with Olsson as last author.
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 and identity resolution come from an OpenAlex snapshot; the h-index (7) and i10-index (6) are raw counts, not field-normalized figures.
- Year-by-year counts list only 2004, 2006, 2007, 2022, and 2026; intervening years are absent from the snapshot and may be incomplete rather than truly zero.
- No ORCID is on file, so this profile is name-resolved from public bibliographic data and could conflate researchers sharing the same name.
- The output and impact axes are unscored because per-paper citation totals and normalized impact aggregates were not available in the input data.