Research on Application of Embedded AI in Fall Detection
Yan Liu, Zhu Xing · 2024
The traditional elderly fall detection system has some shortcomings, such as high false alarm rate and large delay, which limits its reliability and practicability in practical application. In order to timely and accurately notify the guardian of the old man's fall information, a low-power fall detection system based on em-bedded artificial intelligence is proposed. The system is based on low-power ARM processor, coupled with low-power and small-volume MEMS acceleration sensor. By designing the overall structure of the system, the peripheral circuits of ARM processor and MEMS acceleration sensor, and writing embedded programs and embedded artificial intelligence algorithms, the core functions such as data acquisition, fall identification and fall state transmission are realized. In the experiment, an algorithm model of FDNet (Fall Detection Network) is proposed, and a lot of tests and evaluations are carried out. The experimental results show that the system can achieve 98% accuracy on embedded devices. In addition, the system also shows excellent environmental adaptability, low power consumption and small size, which makes it suitable for fall detection and early warning in various occasions. The successful application of the system is expected to provide effective protection for the health and safety of the elderly.