# USC's smart ring measures stress using near-infrared light

> **TL;DR:** The new method, called optodermal activity (ODA), measures sweat activity beneath the skin using two near-infrared LEDs In a test with 6 people, the ring distinguished stress from relaxation with 92.3% accuracy, versus 77.95% for a gel EDA sensor The AI model is just 27.65 kB and runs entirely on the ring, cutting latency to about 33 ms The team plans a 1-year trial with 40 people, and the technology has a patent pending No commercial product or pricing has been announced

- Original: https://migovi.com/en/nhan-thong-minh-do-cang-thang-usc/
- Author: [Sky Shepard](https://migovi.com/en/author/batman11689/)
- Published: 2026-11-22T19:04:50+07:00
- Categories: AI, News, Technology
- Source: [Digital Trends](https://www.digitaltrends.com/wearables/usc-ai-stress-ring/)

---

USC's stress-sensing smart ring runs AI directly on the ring, so sensitive biometric data never has to leave the device for a phone or the cloud.

## Why current stress measurement is inconvenient

Wearables like the Fitbit Sense mostly measure stress using electrodermal activity (EDA), which tracks changes in the skin's electrical conductance caused by sweat gland activity. This approach needs electrodes in steady contact with the skin, so each time you want a reading, you have to rest your hand on the device and hold still for a few minutes. As a result, EDA can't be tracked continuously the way heart rate can, and users have to pause what they're doing whenever they want to know how stressed they are.

A team at the University of Southern California (USC), led by assistant professor Yasser Khan with PhD student Farhad Hassan, has built an AI-equipped ring that takes a different approach: it uses near-infrared light to monitor sweat activity beneath the skin. That lets the ring measure continuously, like a heart rate sensor. You simply wear it, with no extra steps, and the analysis is compact enough to run on the ring itself. The findings were published in the journal Science Advances.

![The stress-sensing smart ring from the USC Khan Lab](https://migovi.com/wp-content/uploads/2026/11/nhan-thong-minh-do-cang-thang-usc-1.jpg)

*The stress-sensing ring developed by the USC team. Photo: USC Khan Lab*

## Two LEDs instead of electrodes

Under stress, the sweat glands activate and the skin beneath changes subtly, even before any sweat appears on the surface. According to Khan, the idea is fairly simple: use light to see that tiny layer of sweat. The ring has two near-infrared LEDs. The 960 nm LED responds to sweat-related changes in water content, while the 1250 nm LED serves as a stable reference. The team calls this method optodermal activity (ODA).

In a test with 4 participants, the ring's optical signal matched a commercial gel EDA sensor with an average correlation of 86% over time. The same optical system also records heart rate and heart rate variability, while a temperature sensor and an accelerometer help the ring tell psychological stress apart from movement or sweating due to heat. Khan says this is the first demonstration that stress can be measured with light.

![Diagram of how the ODA smart ring measures stress](https://migovi.com/wp-content/uploads/2026/11/nhan-thong-minh-do-cang-thang-usc-2.jpg)

*How the ODA ring works. Image: Science Advances / USC*

## Test results and next steps

In a short trial with 6 people, the ring correctly distinguished stress from relaxation in 92.3% of cases, while the gel EDA sensor reached only 77.95%. Across a wider range of conditions, including relaxation, exercise, and heat-induced sweating, the ring achieved 100% accuracy versus 74.58% for EDA. After 12 hours of continuous wear by 2 people, the gel sensor's accuracy dropped to 57.35%, while the ODA ring held at 79.55%. The number of participants is still very small, so these figures are only suggestive.

Khan plans a 1-year follow-up study with 40 people in an office setting. The team is also looking for ways to reduce interference when heavy sweat accumulates on the skin's surface, and is testing more types of cognitive and psychosocial stress. The technology is in the patent application process and could later be licensed commercially through USC's Stevens Center for Innovation.

## AI that runs on the ring

Notably, the ring uses common, readily available components. Temperature sensors and accelerometers are already found in many mainstream devices such as the RingConn Gen 3, so the difference lies in the light source: the team added a 960 nm LED to detect stress and a 1250 nm LED as a reference, along with two photodiodes, whereas commercial rings typically use LEDs in other wavelength bands. From the same optical module, the ring also reads pulse amplitude (PA), heart rate (HR), and their variability (PAV and HRV).

After INT8 quantization, the machine learning model takes up just 27.65 kB. During simultaneous sensing and inference, it needs 174.47 kB of program flash and 27.63 kB of static RAM, small enough to run directly on the ring's microcontroller. This is edge AI: the ring processes the sensor data itself and sends only the conclusion, such as stressed or relaxed, to the phone, rather than transmitting all the raw data. According to the paper, post-measurement latency drops by roughly 300 times, from 9.9 seconds (sending raw data over Bluetooth) to about 33 ms, and battery life under continuous use rises from about 4.2 hours to 6 hours.

![The ODA smart ring platform for stress sensing](https://migovi.com/wp-content/uploads/2026/11/nhan-thong-minh-do-cang-thang-usc-3.jpg)

*The ODA ring platform. Image: Science Advances / USC*

Sensitive biometric data doesn't have to leave the ring, a significant privacy advantage as health data is increasingly pushed to the cloud. The team also says the optical approach isn't limited to rings and could be applied to watches and patches. If a later version moved processing to a phone, a model under 1 MB would be almost negligible, since today's high-end phones can already run AI models with several billion parameters locally.

**MIGOVI VIEWPOINT**

Migovi finds this approach more worth watching than the many current "stress-tracking" rings, which mostly infer stress from heart rate. ODA relies on a more direct physiological signal, and because the hardware uses off-the-shelf components, a real product isn't far-fetched. That said, the best figures come from very small test groups (2 to 6 people), and the ring is still affected when heavy sweat pools on the skin's surface, so this accuracy can't be considered certain in real-world use. For readers in Vietnam, there is no product or price to consider yet. The milestone to wait for is the results of the 1-year, 40-person office trial, since that is much closer to how you would actually wear a ring every day. In the meantime, stress scores on today's wearables should be treated as a rough reference rather than a health conclusion.

Source: [Digital Trends](https://www.digitaltrends.com/wearables/usc-ai-stress-ring/)
