Optical Ring Classified Stress at 92.3%, but Only Six People Were Tested
A research ring used two infrared wavelengths instead of skin electrodes to sense sweat-gland activity, reaching 92.3% binary stress-classification accuracy in a six-person experiment and 87.53% with on-device inference. The prototype may reduce electrode drift, but the cohort was tiny and the study does not establish medical-grade stress diagnosis.
A wearable ring has measured stress-related skin activity with light instead of the metal electrodes used by many existing devices. In a small human study, the prototype combined optical, temperature and motion signals to classify two states with 92.3% accuracy, while a neural network running on the ring's electronics reached 87.53%.
The 30-second summary
- What happened? A USC-led team built a ring that shines 960 and 1250 nanometer light into the finger and measures reflection changes linked to perspiration.
- Why does it matter? Optical sensing could avoid the contact degradation and signal drift that limit electrode-based electrodermal monitoring during long wear.
- What is the catch? The main short-term classification experiment involved six people, and physiological arousal is not a clinical diagnosis of psychological stress.
Key Number: 92.3% binary accuracy. The multimodal optical model reached that result after adding temperature and motion data, compared with 77.95% for electrode-based electrodermal activity in the same short-term experiment.
Watching water change beneath the skin
Electrodermal activity measures changes in skin conductance caused mainly by sweat-gland activation. It is widely used as a marker of sympathetic nervous-system arousal, but conventional sensors require electrodes to remain in stable contact with the skin. Drying gel, movement and changing contact pressure can gradually alter the signal.
The new method, called optical dermal activity, uses a modified light-reflection model. At 960 nanometers, reflected light changes with water concentration in tissue. A second wavelength at 1250 nanometers acts as a reference because it is less sensitive to those changes under the tested conditions. Subtracting the signals helps isolate perspiration dynamics from surrounding tissue absorption.
The ring also records photoplethysmography, the optical pulse signal used by many watches, plus skin temperature and acceleration. That gives the classifier several views of the wearer's physical state rather than asking one sensor to separate stress, exercise and heat-related sweating alone.
What the experiments showed
In a cold-water immersion test involving four participants, the optical signal and electrode-based electrodermal activity had an average correlation of about 86%. The researchers reported similar optical responses across participants with light, brown and dark skin tones, although four people are far too few to establish performance across skin pigmentation.
For the short-term classification experiment, six participants completed conditions representing stress, relaxation, movement and thermal sweating. Optical dermal activity alone produced binary accuracy close to the electrode signal, while the combined optical, temperature and motion model reached 92.3%. In the four-class task, the multimodal system reached 100% within the study data, versus 74.58% for electrode-based sensing.
A one-dimensional convolutional neural network running on the device achieved 87.53% accuracy, showing that inference did not require sending every raw signal to an external computer. The team also reported stable operation during changing light, temperature and humidity and during 14 days of daily wear.
A different measurement, not a stress detector
The prototype fits a wider effort to move wearable health sensing beyond pulse counting. NewTqnia recently covered smartwatch signals used to screen for hypertension and a wireless skin sensor that measures tissue mechanics. The new ring instead targets sweat-gland activation without electrodes.
That distinction matters because perspiration-related arousal can rise during exercise, heat, pain, excitement or anxiety. The classifier can separate the conditions included in its experiment; it does not read a person's mental state independently of context.
Before we overstate the result
- The main short-term experiment included six participants, and the optical-to-electrode correlation test included four.
- The reported accuracies were obtained in structured laboratory conditions, not an independent prospective study of everyday users.
- Skin pigmentation, age, medical conditions, hydration and environment require much broader evaluation.
- The system measures physiological arousal associated with stress. It has not been validated to diagnose anxiety, chronic stress or another medical condition.
- Four-class accuracy of 100% in a tiny dataset is a warning to seek external replication, not evidence of perfect real-world classification.
What must happen next
A meaningful follow-up would train the model on one large, diverse cohort and test it prospectively on different people during unscripted daily life. Researchers would need to report false alerts, missing data, battery life and performance across months, then compare the ring with validated clinical and psychological measures. Until that happens, the work demonstrates a promising electrode-free signal, not a consumer stress monitor ready for health decisions.
Verified topics and entities
Sources and citations3 sources
External references used to support the reporting in this article.
Published by
NewTqnia Health Desk
An institutional editorial team within NewTqnia