Fall Detection with CNN-Casual LSTM Network

Falls are one of the main causes elderly injuries. If faller can be found in time, further injury effectively avoided. In order to protect personal privacy and improve accuracy fall detection, this paper proposes a detection algorithm using CNN-Casual LSTM network based on three-axis acceleration rotation angular velocity sensors. The neural system includes an encoding layer, decoding ResNet18 classifier. Furthermore, layer three layers CNN Casual LSTM. deconvolution maps spatio-temporal information hidden variable output that is more conducive relative work classification network, which classified by ResNet18. Moreover, we used public data set SisFall evaluate performance algorithm. results experiments show has high up 99.79%.