Fall Detection Algorithm Based on MPU6050 and Long-Term Short-Term Memory network
Sheng-Ta Hsieh, Chun‐Ling Lin · 2020
Fall is the second cause of accidental death in the world and is the main cause of physical injuries, especially older. Development of fall detection systems can re-duce the injuries by falling. In this study, three-axis acceleration, three-axis angular acceleration, and Euler parameters that are obtained by MPU6050 sensor are adopted to collect standing, walking, and falling data. Two nrf52832 Bluetooth modules are the receiver and transmitter individual. One reads the data and calcu-lates behavior's information from MPU6050, then transmits data to another on the development board (Nordic nRF52832). Then the development board trans-mits data to the computer through UART. Long-Term Short-Term Memory net-work (LSTM) is used to identify the three movements and then distinguish the difference between falling and normal activities. The results show that this meth-od has 97% rate to determine when fall has occurred.