Fall detection based on accelerometer and gyroscope using back propagation

Abstract

Falling is an external aspect that can lead to death for the elderly. With so many activities they can do will increase the likelihood of falling. A fall detection device is designed to minimize post-fall risk. An MPU6050 sensor with 3 axis accelerometer and 3 gyroscope axis is used to detect the activities of the elderly. This research is expected to recognize the falling forward movement, falling aside, falling backward, sitting, sleeping, squatting, upstairs, down stairs and praying. The total data in the test is 80 data per movement of 16 participants. Backpropagation method is used for motion recognition. The recognition of this movement is based on 10 input variables from the accelerometer sensor data and gyroscope sensor. The result of this study, the error value calculated by using the formula Sum Square Error of all movements, is 0.1818 with ROC accuracy of 98.182%