Spondon

A US$46 multi-sensor wearable for stress detection (ACM COMPASS 2026)

Spondon is a low-cost wearable that detects acute stress from physiological signals, designed for resource-constrained settings where commercial devices (US$300–1,700) are out of reach. It is built from locally available components for about US$46.

Device. An ESP32 microcontroller with a MAX30102 PPG sensor (heart rate and HRV), a GSR sensor (skin conductance), a GY-906 infrared temperature sensor and an ADXL335 accelerometer, on a custom PCB. Data streams over Wi-Fi to a custom Android app.

Study. 15 participants completed a 5-minute baseline, a 5-minute Stroop task and a 5-minute recovery period (NSU IRB-approved).

Findings.

  • Raw signals generalize poorly to new people (47.9% accuracy), because resting physiology varies widely between individuals.
  • Expressing each signal as a change from the person’s own baseline raises leave-one-subject-out accuracy to 75.0% for binary stress detection with a Random Forest.
  • Three-class classification (baseline, stress, recovery) reaches 58.4%, against a 33.3% chance level.
  • Skin conductance rose significantly during the task (p < 0.001) and was the strongest predictor.
Left: skin conductance relative to each participant's baseline, rising during the Stroop task (shaded). Right: leave-one-subject-out accuracy by model.

Links: Paper (ACM Digital Library) · Analysis code

Shahariar Ifti, Rahat Jahangir Rony, Shuvashish Chakraborty, Asif Mahbub, Nova Ahmed. ACM COMPASS 2026, short paper.