Authors: Ms. Gauri J. Sutar, Associate Professor Dr. Sarita V. Balshetwar, Assistant Professor Mrs. Rajani M. Mandhare
Abstract: This paper presents a low-cost pediatric wearable wristband for continuous fever-risk monitoring using temperature, heart-rate, and motion sensing. The device integrates an ESP32 controller, a DS18B20 temperature sensor, a pulse sensor, an MPU6050 accelerometer, a GSM module, a local buzzer/LED alert, and a Blynk cloud dashboard. Because the wrist skin temperature differs from core body temperature, the temperature channel is calibrated against a clinical digital thermometer and reported as a calibrated body-equivalent temperature. A lightweight, transparent Fever Risk Score (FRS) combines physiological and motion parameters, and a supervised machine-learning classifier (Random Forest) trained on a labeled dataset of 2,400 pediatric monitoring instances classifies the child’s condition into Normal, Warning, and High-Risk states. On a held-out test set the classifier achieved 91.4% accuracy, 94.5% sensitivity, and 98.1% specificity for High-Risk detection (macro F1 = 0.90). Bench validation gave a mean absolute temperature error of 0.18 °C against a clinical thermometer and 2.45 bpm against a pulse oximeter, with 95.5% fall-detection accuracy. Dual-channel alerting delivered GSM SMS in 6.5 ± 1.2 s and Blynk cloud updates in 1.9 ± 0.6 s, and the prototype operated continuously for 18.6 h on a 3.7 V Li-ion battery. The results indicate a validated, deployable early-warning and caregiver-alerting tool suitable for homes, schools, and rural healthcare settings, positioned as decision support rather than a replacement for clinical diagnosis.
