Development Of Wearable Fall Detection And Protection System

Fall detection research is limited to laboratory precision equipment,while existing simple bracelet or watch can only obtain simple and low-precision fall information.Therefore,it is of great significance to develop wearable real-time fall detection algorithm and provide effective protection to reduce the damage caused by falls in the elderly.In this paper,two fall detection algorithms,convolutional neural network algorithm and threshold algorithm,are proposed.The advantages and disadvantages of the two algorithms are compared.The threshold algorithm is implemented on the single-chip microcomputer.Due to the large error in the value obtained by the acceleration integral,the obtained velocity and displacement were verified by the optical capture system.The protection system used in this paper controls the trigger mechanism through the IO port of the single-chip microcomputer to open the high pressure gas cylinder and make the compressed carbon dioxide fill the air bag in a very short time.Considering its safety and environmental protection,a mechanical trigger mechanism is designed.With the innovative help of Qin crossbow design,the movement of the suspension knife is controlled by controlling the rotation of the steering gear.After that,the spring is released,and the cylinder is punctured by an external force.The feasibility of the mechanical structure is verified theoretically through force analysis and finite element analysis of the structure.A wearable real-time fall detection and protection system is further developed.The system adopts modular design to realize the collection,transmission and calculation of human motion data and output control signals in real time.In this paper,one inertial sensor was used and placed around the waist.Through the experimental verification under the optical capture system,in outdoor,the performance of the system's fall detection,misjudgment of daily behavior and timely trigger of the air bag was good.The sensitivity and specificity of this algorithm were 96.7%and 91.5%,respectively,by setting a threshold value on the time series to determine the fall,and it could trigger 363 ± 174ms before the end of the fall.The average trigger time of the mechanical trigger structure was 0.27s,meeting the application requirements.The innovation of this paper is that the velocity and displacement characteristics are obtained by fitting the trend with lag filtering,which can eliminate the error of acceleration integral greatly.The displacement feature is introduced into the threshold algorithm,which eliminates the misjudgment of daily behaviors such as jogging and jumping.The designed mechanical trigger mechanism is small in size,large in force magnification,easy to control,and suitable for protection systems.