Similarity ratings differed by scent and material
A small preprint evaluation of ScentEcho found that people’s ratings of replayed odors varied with both the scent and the adsorbent material used to capture it. On a 1-to-7 scale, material A averaged 4.5 for mint, while material D averaged 5.3 for lavender. Coffee scores were tightly grouped at 4.2 to 4.3 across materials, and lemon scores ranged from 3.9 to 4.1.
Intensity ratings also differed among the material and odor combinations. Material D scored 5 for lavender, while materials A, B and C each scored above 4 for coffee. Similarity and intensity were both rated relative to the original target on a 1-to-7 scale, with the target defined as 7.
The two ratings were closely linked
The relationship between the two scores was strong in the reported analysis. Using a Pearson correlation, a measure of whether two sets of scores move together, the study reported a coefficient of 0.8295 for similarity and intensity. Participants generally gave higher similarity ratings when they perceived replayed odors as stronger. The association does not establish that increasing intensity by itself causes a closer match.
Participant reports added a qualification. Seven of the 12 said they had difficulty separating similarity from intensity. The remaining five said they could evaluate the two dimensions separately when intensity was sufficiently strong. These were self-reported judgments, and the evaluation did not quantify a formal intensity threshold for that separation.
Recognition appeared to follow a graded pattern
The evaluation also found an exploratory recognition pattern. Participants tended to identify broad odor categories at similarity scores of 3 to 4, while scores of 5 to 6 were associated with more specific odor sources. The report describes this as a pattern in participant reports rather than a validated recognition analysis, so the score bands should not be treated as universal recognition thresholds.
Separate qualitative interviews asked how participants experienced odor fidelity and formed expectations about reproducing everyday scents.
What the evaluation tested
ScentEcho has separate upper control and lower storage compartments, with four independent odor storage units.
For this evaluation, it collected lemon, coffee beans, mint and lavender for one minute on four adsorbent materials labeled A through D, then replayed the odors. Participants rated each replay for similarity and intensity on a 1-to-7 scale, using the original target odor as the 7-point reference. A three-minute rest was provided between target odors. The report does not say whether the order was randomized or participants were blinded, and it does not report an independent control condition, preregistration or a power calculation.
Results came from 12 participants, six males and six females, aged 22 to 49, with an average age of 31.2 years and a standard deviation of 7.3. The reported scores describe this group’s responses to the selected replayed odors.
Promising results, practical shortcomings
Several material and odor combinations received similarity scores above 5, which the authors interpreted as high perceptual fidelity. The report does not fully list every combination above that level, and performance differed by odor, so the findings do not identify one material as best for all four scents.
The authors also report that replayed odors were generally weaker than the originals. Some participants noticed faint metallic smells with weaker samples, and the report notes that the pump may continue operating when all solenoid valves are closed.
This is an arXiv version 1 preprint dated 26 Aug 2026. The work was supported by the Foundation of the National Key Laboratory of Human Factors Engineering and the National Natural Science Foundation of China through the reported grants.
Paper data and sources
Original title: ScentEcho: Exploring Adsorbent Materials for Accurate Odor Collection and Playback
Authors: Chih-Hung Lee, Yuchi Sun, Rui Zhang et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text