Authors: Dnyaneshwari Babasaheb Niwdunge, Rasika Jijabhau Gunjal, Priti Vilas Hadawale, Dr. V. M. Dhede, Prof. M. S. Nalawade
Abstract: Many personal safety devices depend on the user pressing a panic button or unlocking a smartphone, yet in a fall, a sudden medical event, an assault or harassment the person may be unable to do either. A wearable that recognises such situations by itself and calls for help silently can close this gap. This review examines the technologies needed for such a device and organises them around SilentSOS, a multimodal context-aware wearable concept. It surveys physiological sensing of heart rate and blood oxygen saturation by photoplethysmography, motion sensing and fall detection with inertial sensors, skin and ambient temperature monitoring, acoustic analysis of surrounding sounds, satellite positioning and cellular alerting. It then reviews how machine learning combines these signals so that the decision does not rest on a single sensor, which reduces both missed events and false alarms. A layered architecture is presented in which an ESP32 microcontroller acquires data from MAX30102, MPU6050, temperature and microphone sensors, extracts features, classifies the situation as normal, suspicious or high-risk, and sends an SMS with location and readings through a GSM module, with a discreet gesture available for manual activation. Design trade-offs including battery life, privacy, on-device inference and validation are discussed, together with future extensions such as ECG and blood-pressure estimation.
DOI: https://doi.org/10.5281/zenodo.23863454
