Long short-term memory networks for bed position classification
Sakada Sao · NRCT Data Center · 2019
This paper describes an approach of 2 stacked layers of Long Short-Term Memory for bed position classification. The sensor panel data is collected from 2 types of sensors under the 30Hz sampling rate, i.e. piezoelectric and pressure sensors. The data is classified into 5 classes and those data are normalized with min-max scaling on a fixed range between 0 and 1. The model is experimented using window sliding technique for preprocessing dataset with changing the number of hidden nodes in 128, 80 and 50 nodes of the model. The overall accuracy is 94.74% which can be comparable with the previous work.