Project

Saving Face

Zhi Wei Gan,  Irmandy Wicaksono

“Don’t touch your face” is seemingly simple advice. Since coronaviruses are stable for days on many surfaces, a person can get COVID-19 by touching a contaminated handle or object and then touching their own mouth, nose, or possibly eyes. But quitting is far easier said than done. Most people touch their face frequently throughout the day, usually without thinking about it—it’s a very difficult habit to break and requires a surprising amount of conscious effort.

The MIT Media Lab is advancing Saving Face: a suite of easily scaled technologies to help people fight the pandemic by warning them when they’re about to touch their faces. 

Responding to a conceptual challenge launched by Professor Kevin Esvelt, head of the Sculpting Evolution research group, researchers from the Media Lab community came up with three classes of sensing technologies capable of detecting hand-face proximity and alerting the wearer. 

In “record” mode, users can train themselves to break the habit by learning how often they touch their face, while “alert” mode will directly warn them off with an alarm or vibration whenever they make an att… View full description

“Don’t touch your face” is seemingly simple advice. Since coronaviruses are stable for days on many surfaces, a person can get COVID-19 by touching a contaminated handle or object and then touching their own mouth, nose, or possibly eyes. But quitting is far easier said than done. Most people touch their face frequently throughout the day, usually without thinking about it—it’s a very difficult habit to break and requires a surprising amount of conscious effort.

The MIT Media Lab is advancing Saving Face: a suite of easily scaled technologies to help people fight the pandemic by warning them when they’re about to touch their faces. 

Responding to a conceptual challenge launched by Professor Kevin Esvelt, head of the Sculpting Evolution research group, researchers from the Media Lab community came up with three classes of sensing technologies capable of detecting hand-face proximity and alerting the wearer. 

In “record” mode, users can train themselves to break the habit by learning how often they touch their face, while “alert” mode will directly warn them off with an alarm or vibration whenever they make an attempt. 

All three sensing techniques have demonstrated promise in concept studies. Given the urgent need to develop and make these training devices available to as many people as possible, the nonprofit team decided to pursue the three technologies in parallel.

The approaches include:

  1. Detecting the distance between a user's smartwatch and earbuds using the strength of the Bluetooth LE signal they use to communicate (RSSI) and alerting the user that their hand is approaching their face. A pre-alpha app can currently record total face touches, while alert mode is under development.
  2. Using a SONAR-inspired approach to measure the distance between hands and face and warning the user with an alarm when they get too close. This technique transmits an ultrasound signal from earbuds worn on the wrists and receives it with a microphone near the face. The team confirmed that many inexpensive off-the-shelf wired earbuds can generate and detect 20kHz ultrasonic frequencies, so this solution could in principle be used by anyone with a smartphone for less than $5 in headphones and cables. An app offering both record and alert modes is in development.
  3. Building a device to sound a buzzer when the hand and face get too close, as detected using electromagnetic or capacitive fields in one of two ways:
    1. An electric field generated by conductive fabrics/stickers worn around the neck is altered by hand movements toward the face, or
    2. A magnetic ring or bracelet is detected by a magnetic sensor worn on the head or the neck.

The team has determined that all the above methods can be sensitive enough to detect a hand 20 centimeters from the face, which is far enough to deliver a timely warning in "alert" mode. The team is currently developing apps compatible with off-the-shelf hardware, designing techniques to cheaply produce new analog devices using electromagnetic field sensing, and testing these approaches for robustness and efficiency.

Ideas were generated and advanced by a task force including Camilo Rojas (postdoctoral researcher, Fluid Interfaces), Irmandy Wicaksono  (research assistant, Responsive Environments), Eyal Perry (research assistant, Molecular Machines), Cedric Honnet (visiting scientist, Responsive Environments), Niels Poulsen (visiting student, Fluid Interfaces) and Zhi Wei Gan (UROP, Opera of the Future), along with professors Kevin Esvelt (Sculpting Evolution), Joe Paradiso (Responsive Environments), and Fadel Adib (Signal Kinetics). An initial proof-of-concept prototype using Bluetooth low-energy on iOS was written in partnership with a startup called Augmental Tech launched by Media Lab alum Tomás Vega, who donated time to the nonprofit effort. 

To advance the development and rapid deployment of a solution that can be made widely available, the team is now engaging in collaborations with leading tech companies. They aim to make the alpha version of at least one open-source app available next week.

Interested in staying informed? Share your smartphone OS, preference for smartwatch vs earbuds, and sign up to receive notifications of new releases.

For questions or ways to help out, please contact donttouchyourface@media.mit.edu.