Edge-IoT Facilitated Multimodal Sensor Fusion for Threat Recognition in Women's Safety Wearables

Safety of women is a significant problem on the
global level, and the fact that the time gap between the threat appearance and its detection can be lowered is one of the pressing social problems. Smart wearable devices coupled with smart systems are being explored more and more with this reason. Conventional safety systems that use only one sensorlike accelerometers or GPS are usually characterized by false alarms, low accuracy and slow responses. The paper presents an Edge-IoT-based multimodal sensor fusion model of female safety wearables in order to overcome these drawbacks. The framework combines the signal of accelerometer and gyroscopes and conducts on-device inference to make real-time responsive and communicate with the cloud to support alerts. The experimental validation with the UCI Human Activity Recognition dataset had shown a classification accuracy of 93 percent, lower latency, previously 7.8 s (cloud only) to 3.2 s
(edge), and a reasonable battery consumption (88 percent). The findings validate the assertion that multimodal sensor fusion yields consistency and precision in accuracy, responsiveness and reliability when compared to single-sensor systems. It will be extended in the future with physiological and acoustic sensors to enhance the detection of a context-based threat, accompanied by studies in the real world to assess userfriendliness, ethical standards, and energy efficiency.