Fall Detection using Deep Learning with Sensing Data

초록

The construction industry plays a significant role in a nation's development. However, construction industry workers face high risks due to the use of heavy machinery and require constant vigilance to prevent accidents, such as falls. Sensor-based fall detection systems are considered accurate and efficient as they are easy to carry when attached to the body. In this study, the data is collected while performing some regular tasks using wearable sensors, the gyro and accelerometer. The collected data was then processed and cleaned after which the learning model was applied to it. CNN-LSTM-based model for detection of fall was introduced and applied together because of CNN feature extraction capability along with LSTM as it is popular for its time series ability. The proposed CNN-LSTM-based model, using an accelerometer and gyro sensor at a sampling frequency of 50 Hz, has an accuracy of 99.91%.

키워드