The Design And Implementation Of Fall Alarm System Based On Machine Learning

Taking care of the elderly is a problem which all countries in the world is confronted,especially for the rising aging country,the problem is particularly prominent.How to alert in time when the elderly is falling down is the important measure to protect the safety of their lives.With the rapid development of information technology,the use of wearable device,such as smart phones,provides favorable conditions to monitor the fall of the elderly with real-time,however,the portability of the related equipment is limited,the reliability is not enough,and the accuracy is not high.This thesis uses android phone as the platform by collecting the real-time acceleration and combining the relevant machine learning algorithms to realize fall alert system.First we put forward an improved feature selection algorithm for data preprocessing,and get a feature subset,and then puts forward a three-level hybrid fall detection algorithm,use the selected feature to make judgement about falling and alert.This system mainly contains fall alert module,fall detect module,real-time position module,the main idea of this system is to judge the falling correctly and send message with position information in time.In order to verify the reliability of fall alert system,a variety of experiments have been carried out to verify falling state and not falling state.By experimental verification,the accuracy of falling state is above 90%,the rate of false positives of not fall state within 10%,and the distress message can be sent to the designed mobile phone,thus proving the reliability of this system is high,and can reach the function of fall alert.