Digitising Smell: From Natural Statistics of Olfactory Perceptual Space to Digital Transmission of Odors

Projektinformationen

D2Smell

ID Finanzhilfevereinbarung: 101118977

DOI

10.3030/101118977

EK-Unterschriftsdatum 29 Januar 2024

Startdatum 1 Februar 2024

Enddatum 31 Januar 2030

Finanziert unter

• European Research Council (ERC)

Gesamtkosten

€ 11 835 892,00

EU-Beitrag

€ 11 835 892,00

Koordiniert durch

WEIZMANN INSTITUTE OF SCIENCE
Israel

Periodic Reporting for period 1 - D2Smell (Digitising Smell: From Natural Statistics of Olfactory Perceptual Space to Digital Transmission of Odors)

Berichtszeitraum: 2024-02-01 bis 2025-07-31

Zusammenfassung vom Kontext und den Gesamtzielen des Projekts

We aim to digitize smell. Achieving this is currently prevented by gaps in basic science. We aim to fill these gaps, culminating in a proof of concept for our model. The primary gap we identify is lack of data on what humans typically smell. Phrased conceptually, in Aim 1 we ask what are the natural statistics of human olfactory perceptual space. We address this in a series of three experiments, highlighted by one where we equip participants with a wearable sampling apparatus we designed and built for this proposal. The apparatus measures sniffing behaviour to identify odor sampling, measures neural activity to verify olfactory perception, takes video of the visual scene, analyses total levels of volatile organic compounds in real time, and collects odorant samples for detailed analysis off line. In other words, we generate an olfactory equivalent of Google Street View, with the addition of chemical, perceptual and neural data. Using this we will characterise the natural statistics of human olfactory perceptual space. Moreover, a major contribution of this proposal will be in posting this massive data as a publicly available recourse. Next, in Aim 2 we build on this data to digitize human olfactory perceptual space. We put forth a model that allows us to recreate odors using a restricted set of odor primaries. We will test our model in two frameworks: One we call SmelloVision, where we develop the algorithmic framework to generate an odor to match digital images, and one we call TelleSmell, where we use a device to sense the environment, an algorithmic framework to transfer the data, and a device to generate the corresponding odor remotely.

Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse

The D2Smell project entails first building significant physical and computational tools, and then using them to amass data and answer questions. These tools have been put in place: We launched the SMELLiT platform to collect olfactory perceptual entries in mass. SMELLiT is now available on the Android (https://play.google.com/store/apps/details?id=com.weizmann.worg.smellitapp&hl)
and IOS (https://apps.apple.com/il/app/smellit/id6502876469) platforms, and have already amassed thousands of entries. We also launched the “what we smell database” (WWSdb) (https://data.d2smell.org/form/sample) a public resource containing the chemical contents of common odors. Moreover, we have completed the purchasing and instalment of several analytical platforms for odor emission and analysis (InfrustructureFigure), and have started collecting relevant data.

Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)

The IBM-hosted DREAM CHALLENGES held a competition for predicting odor perception from structure - a critical step in the digitisation of smell. We participated in this competition as a group (under the group name D2Smell), and are happy to report that our team tied for first place in this competition (https://www.synapse.org/Synapse:syn53470621/wiki/626022) thus further validating our computational approach. Moreover, using the new online tools we have developed, we now have a first glimpse at what people smell. In a word - its coffee :-), and other foods, that together account for a quarter of what people smell. Remarkably, humans are aware of an odor for at least 15% of their wake time, and in contrast to popular notions, men are more aware of odors than women (ResultsFigure). Moreover, in the first direct publication resulting from this project, we developed models aimed at detecting an odor perception event from EEG data - a critical tool for the continuation of this project. All together, we have put the necessary tools in place, and are confident that in the second half of this project we will use these tools to generate the intended data and outcome - digitisation of smell.

Infrastructure Figure

Results Figure