Low-Power Fall Detection System with Location Versatility
Tao Xu, Jiahui Liu, Manghe Geng · 2022 41st Chinese Control Conference (CCC) · 2022
In response to the low power consumption and comfortable wearing requirements of wearable fall detection technology, a fall detection algorithm triggered by interrupt was proposed in this paper. The acceleration and angular velocity measured by MPU6050 were utilized to track different stages of the fall, which achieved the purpose of identifying falls. Software and hardware were combined to reduce power consumption. A dataset composed of 1400 simulated falls and 900 ADLs has been used to train the fall detection algorithm, and it was tested in real scenarios. The results show that the algorithm can be applied in different locations, including chest, back, waist, left and right pockets of pants. The best location is the waist, and the accuracy reaches 100%. At the same time, the best sensitivity, best specificity and best accuracy achieve by the overall location are 100%, 100%, and 99.43% respectively. In addition, it is estimated that the maximum service life of the equipment is 50 hours. Compared with the existing fall detection system, this system has the advantages of low power consumption, high accuracy, and position versatility.