AN EFFECTIVE THRESHOLD BASED MEASUREMENT TECHNIQUE FOR FALL DETECTION USING SMART DEVICES

Abstract

Falls can be considered as most critical events for human workers in real world scenarios which require timely response from the emergency team. Although many have come up with fall detection devices, complex sensors arrangement and response time remain as the challenges on automatic detection, particularly in industrial environment. This paper proposes an effective fall detection algorithm using threshold based measurement approach that consists of two stages. The first focuses on optimizing the thresholds from the wearable sensor data and is required to run only one time for a specific device. The second proposes fall detection algorithms using inertial units and orientation sensor from smart devices to detect the fall. The proposed algorithms in this study take into account accelerometer and gyroscope sensors for fall detection and an orientation sensor to validate the detected fall. The wearable sensors used in this study are very common and thus does not require any special arrangement to wear them. 30% of the fall simulation data was used to acquire the optimized thresholds whereas 70% of it was used for testing of the proposed algorithm with optimized thresholds. The experiment results show better trade-off in terms of sensitivity, specificity and detection time, in comparison to the existing studies. This study also provides experimental study of fall detection algorithm by changing the placement of sensors to three different locations. It indicates the efficacy of the proposed algorithm and can adapt to changes of smart devices.