Neural network modelling of daily life and fall activity for sensor detection system.

Falling is a well-known cause of injured among elderly people. It always acted as a threat toward elder people. Fall detection system is one of the methods for solving the falling issue which occurred in daily life of elder. A model based is essential to identify the behavior of daily life or fall activity. Thus, this study presented a modelling for fall detection system by using neural network algorithm which we had known as nonlinear autoregression, NARnet. The algorithms which will be used in simulated the model are training Levenberg-Marquardt (LM), Scaled Conjugate Gradient (SCG) and Broyden–Fletcher–Goldfarb–Shanno (BFG) function. Four volunteers between 140cm to 170cm height perform eight scenarios. Then, the data set is divided to nine (120LM, 120BFG, 120SCG, 240LM, 240BFG, 240SCG, 1616LM, 1616BFG, and 1616SCG) from 1 to 20 hidden node have been tried and ran in MATLAB. The best model will be selected among nine datasets based on Mean Square Error (MSE) and Regression and evaluated by using decision matric table. From this research, the best model from this research is at 120 data with learning algorithm LM on hidden node 14. [ABSTRACT FROM AUTHOR]