Development and use of intelligent helmets to monitor the health status of emergency workers during rescue operations

Abstract

The article presents the development and pilot testing of an intelligent helmet for continuous monitoring of the health status of EMERCOM personnel during emergency-rescue operations, aiming at early detection of critical conditions and reduction of injuries by shifting from episodic checks to proactive analytics based on multiparametric data. Materials and methods include a prototype built on a standard protective helmet integrating an optical pulse oximeter(HR, SpO2), an in-ear digital thermometer, a three-axis accelerometer and gyroscope (fall/impact detection), an electrochemical CO sensor, and a low-power microcontroller providing on-board preprocessing and LoRaWAN telemetry to a gateway and a server with a time-series database, dashboard, and analytics modules(including an LSTM forecast). The experiment involved 30 volunteers across three scenarios simulating real conditions: high temperature and smoke, debris clearance(high physical load), and a gas-polluted environment(controlled CO); medical validation of measurements was conducted in parallel. Results show statistically significant HRV reductions(SDNN from 138,4±15,2 ms at rest to 39,7±7,1 ms under combined load; RMSSD from 45,6±8,3 to 12,9±3,6 ms), SpO2 drop to 93,4±1,9%(in 15% ≤92%), body temperature increase to 38,6±0,4 °C, and CO elevation to 155,7±23,8 ppm; average alarm latency was 2,5–3,7 s. State classification achieved Accuracy 98.0%, Recall 97,7%, and specificity 98,1%, with a low false-alarm rate(1,9%) and rare missed events(2,3%); an integral functional state index formed using AHP weights for HRV/SpO2/temperature/accelerometry improved predictiveness, and the LSTM provided a 5-7-minute IFS forecast with about 85% accuracy. The discussion notes reduced risks of overexertion, hypoxia, and heat stroke through early trend recognition and supports integrating helmets into EMERCOM’s operational control loop; limitations relate to sample size and simulated conditions, while prospects include miniaturization, extended autonomy, a broader sensor suite, and algorithm personalization.