Research And Implementation Of Human Fall Detection System Based On Motion Sensor

China is one of the countries with the most aging population in the world.More and more elderly people lose their ability to move due to disability and cause serious burdens on families.Among them,accidental falls are one of the major causes of disability in the elderly.The physical function of the elderly is declining,and their resilience is weak.After falling,it can easily cause serious consequences.Many of the casualties caused by the falls do not come from the fall itself,but more often because of the failure to receive timely and effective rescue after the fall.If they can quickly and effectively get rescue after a fall,we can avoid all kinds of bad consequences caused by the failure of rescue.In order to ensure that elderly people to get timely rescue after falling,many scholars at home and abroad have studied the fall detection.There are three studies on fall detection,image-based fall detection,environment-based fall detection,and fall detection based on wearable devices.Among them,the fall detection based on the wearable device has the advantages of being free from the influence of geographical environment,rapid response and wide range of applications compared with other two research methods.The research in this article is based on the motion sensor of an Android mobile phone and can also be classified as a wearable device fall detection.The purpose of the fall detection is to distinguish the fall from daily activities.This article first analyzes the fall and daily activities and defines nine daily activities and four falls.Secondly,the data acquisition plan was designed.By invoking the motion sensor in the Android mobile phone,the acceleration and angular velocity data of various behaviors were recorded,and the 130 groups of samples were analyzed and arranged.Again,SVMs(Support Vector Machines)using different kernel functions are used to train the samples.By contrast,the linear kernel function SVM training model is selected,and the fall recognition algorithm is designed according to the model.Finally,a fall detection algorithm based on the design of the application was designed and implemented.A fall detection system for Android platform based on motion sensor was designed and implemented.After the system detects that the human body has fallen down,it can promptly and effectively call the emergency contact telephone set in advance.At the same time,using mobile positioning,locate the specific location of the fall and send it to the emergency contact mobile phone in the form of a text message.If it is a false alarm,you can manually cancel the alarm.Most fall detection systems use acceleration sensors to capture human acceleration and compare it with pre-set thresholds to determine if a fall has occurred.However,judging a fall merely by the acceleration threshold does not give a very high accuracy.In this paper,the SVM training data samples are used to obtain a classification model,and based on the classification model,an algorithm is designed to determine whether or not a fall has occurred.Finally,the accuracy of the fall detection was verified by the design verification scheme,and the reliability of the fall recognition algorithm was highlighted.Theoretical and experimental results show that the system has high detection accuracy,stable performance,and can be alerted in time after human falls.The system not only has very important significance in protecting the health of the elderly,but also has broad development prospects in the medical field.