You approach a flower on the screen, but its scent arrives a little late. This timing mismatch is a challenge for technology that combines visuals and smell. Even if the scent generator operates quickly, the released molecules still need time to travel to the nose.
A study published by Yiming Liu and colleagues in 2024 combined miniature scent generators with a system that predicts human movement and the spread of scent. The goals were to present scents that fit actions in a virtual environment and to support training with many different odors.
Heating fragrance and rapidly opening and closing the outlet
The scent generator has a layered structure that includes cotton holding fragrance material, a thin heater, a temperature sensor, and a small electromagnetically actuated lid. Temperature control vaporizes the fragrance, while opening and closing the outlet holes controls its release.
Changing the temperature alone takes several seconds to reach the target. The researchers therefore combined temperature control with mechanical opening and closing to switch scent release more quickly. Multiple generators can be operated independently, allowing users to select both the scent and how it is released.
The approximately 70-millisecond response reported in the paper was measured using ethanol: it was the time between closing the outlet and the start of a decrease in concentration near that outlet. It does not mean that people perceived a scent 70 milliseconds after the power was switched on. Nor does it mean that all scent already released into the surrounding space could be removed in that time.
Sending scent ahead to where a person will arrive
In the mixed-reality, or MR, demonstration, scent generators and positioning devices were built into real objects, whose locations were matched to objects in the virtual environment. The devices released scents corresponding to tea, gardenia, osmanthus, and pineapple.
The system uses the participant’s position, facing direction, and movement speed, together with the generators’ locations and environmental conditions such as airflow and temperature. It predicts where a person continuing along their path will enter an area reached by the scent, and begins releasing it in advance, allowing for the time the molecules need to get there.
Another important feature is that the generators operate when a person is predicted to approach, rather than running continuously. The authors propose this as a design that reduces unnecessary fragrance release and power consumption while reducing the mismatch with visuals. In the demonstration, one man moved around four scent-emitting objects.
However, the scent-dispersion calculations use simplifications, including ignoring turbulent airflow and assigning the same perceptual threshold concentration to all four scents. They do not reproduce every difference in sensitivity between people or scents, or all the complexities of indoor airflow. The work has not yet established that many people in a variety of settings would genuinely experience no perceptible delay.
This demonstration also placed the scent generators in the objects. The paper itself states that, because the positioning device is bulky, the same predictive system has not been integrated into the skin-mounted version to eliminate latency.
Switching between 32 scents on the forearm
Separately, the researchers arranged 32 scent generators in a flexible device worn on the forearm. This was used for training in which many different scents were presented one at a time and participants named them.
They also tested whether releasing two scents simultaneously would be perceived as a single new scent. When 10 people were presented with 36 combinations made from nine scents, the proportion perceiving a new scent varied by combination. The authors concluded that mixing did not yet reliably produce a desired new scent, so training used one scent at a time.
The training evaluation compared 11 people who had about an hour of practice beforehand with 14 who did not. Identification accuracy was higher in the practiced group, and repeated trials brought increases in accuracy and shorter response times. These were changes in performance on the task of distinguishing and naming the scents used in this study. They do not establish a broad improvement in sensitivity to faint odors.
Allocating practice time to each individual
The researchers also developed an AI system that adjusts the training time allocated to each scent based on performance. A separate group of 10 people without olfactory impairment used it for seven days and showed improvements in accuracy and response times. The authors see promise in allocating time according to individual needs rather than giving every scent the same amount of time.
The study also reports a trial in one patient who had lost their sense of smell after COVID-19. However, the course of a single case cannot separate the device’s effects from natural recovery or learning through repetition. Establishing clinical treatment efficacy requires testing with an appropriate comparison group.
Mood questionnaires were also administered to the 11 participants who had practiced, but not every measure changed in the same direction. The sample was small, and this study does not establish olfactory training as a treatment for anxiety or depression.
Conclusion
The innovation in this study was to combine small, fast-switching scent generators with calculations that select presentation timing and training content. In MR, predicting both human movement and scent transport was intended to deliver scent ahead of the user’s arrival. The forearm device demonstrated switching between many scents and allocating training time to suit each individual.
The authors envision technology that treats scent as information, with applications ranging from entertainment and education to support for olfactory recovery. Practical use requires addressing not only device response, but also surrounding airflow, scent type, and differences in perception between people. Training applications also remain at a stage where improvement on the task must be distinguished from changes in smell that are useful in everyday life.
References
Liu, Y., Jia, S., Yiu, C. K., Park, W., et al. (2024). Intelligent wearable olfactory interface for latency-free mixed reality and fast olfactory enhancement. Nature Communications, 15, 4474. DOI: 10.1038/s41467-024-48884-z / Full original article.
The research was supported by organizations including Hong Kong’s Research Grants Council, City University of Hong Kong, the National Natural Science Foundation of China, and the Japan Society for the Promotion of Science. The authors declare no competing interests. Original article: © 2024 The Author(s), CC BY 4.0.
