Long Short-Term Memory for Human Fall Detection Based Gamification on Unconstraint Smartphone Position

Abstract

Fall incident can caused health problem in elderly and people with special treatments. Fall detection method is needed to minimized the problem when human fallen and smartphone can be used as the device to detect it. Usually people carry smartphone in any positions and can make fall detection method difficult to detect when fall occurs. Long Short-Term Memory (LSTM) combined with Sigmoid helps to answer the challenge to handle smartphone accelerometer and gyroscope data in many smartphone orientation position. LSTM method on this experiment can achieved 91.67% for the accuracy result. Since the fall data occurs rarely and there may be insufficient data available, the gamification prospect implemented in fall detection application especially on “Bug” bounty can help researcher to enhance the accuracy result to make a better human fall detection method.