Research On Human Fall Detection Algorithm Based On Wearable Health Cloth

Nowadays,with the continuous growth of the elderly population in our country,the degree of social aging has gradually deepened.For the elderly,the physical condition is declining,the ability to maintain body balance is getting worse,fall events usually occur.The fall event has become the biggest threat to the elderly,and the consequences are not only physical trauma,but also death.The fall detection method not only provides the safety protection for the elderly,but ensures the healthy independence of daily life.In the medical environment,the fall detection enables the medical staff to immediately know the patient's fall,thus providing medical assistance timely and quick assessment of the condition.It is also significant for improving the quality of medical care.Therefore,the study of fall detection has a high research and practical significance.At present,the research on the method of fall detection has many problems with high false alarm rate.The method of falling detection is rarely related to the establishment of mathematical model of fall behavior.There are still many key problems not be resolved to design the fall detection method based on mathematical model.In this paper,we explore the fall behavior detection by combining the wearable equipment and the theory related to the establishment of the mathematical model.The main work and contributions of this study are summarized as follows:(1)Design and build a wearable human motion monitoring experimental system.Based on the wearable health clothing developed by the Laboratory of Bioinformatics,Chang Gung University,the system of human motion monitoring experiment was established.The system can collect the information of health clothing,and designed and developed an information processing program on PC.The acceleration signal of daily behavior and fall behavior of the human is collected by the system,and the difference between the two kinds of behaviors acceleration signal is analyzed.(2)Define the fall behavior process.This paper summarizes the characteristics of the falling motion acceleration signal in the amplitude change and the time process,sets the different phases and motion states of the falling acceleration signal.The states of motion can be used to describe the signal of fall events.(3)Propose a fall detection algorithm based on hidden Markov model.Firstly,the observation sequences extraction method is designed to deal with the acceleration signal of the human motion process.The time series of the acceleration of the characteristic is obtained,and the motion behavior is described in the form of "observation time sequences".Secondly,Different movement state changes can sum up the characteristics of fall process during time changes.According to the characteristics of the acceleration of the fall behavior,the unbalanced state and the collision state during the fall process,hidden Markov model describing the fall process of the human body can be trained,and apply the hidden Markov model to calculate matching probability of the model and the observation sequence to assess the probability of fall events.(4)Design an experiment to validate the fall detection algorithm proposed in this paper.The fall behavior model is used to analyze the behavior of the daily behavior and the fall behavior.In the range of experimental verification,the fall detection algorithm based on Hidden Markov model can effectively distinguish the daily behavior and fall events.The algorithm is verified by the 5-fold cross validation.The accuracy rate is 99.1%,the sensitivity is 96.7%,and the specificity is 99.3%.(5)Design and implement an APP about fall detection on mobile The practical application platform of the Android is selected to test the practicability of the method.With the portability of the Android software,this fall detection APP lay the foundation for the later development of the health monitoring software.