Research And Application Of Fall Detection Algorithm Based On Body Posture

With the intensification of aging problem and the extension of expected life span, as well as the frequently appearing of "empty nest family", the elderly issue also become prominent, especially the healthy problem, becoming one of the key issues concern to the country and society. Falling by accident with bring harm to both mental and physiological health, rescue on time will greatly reduce the disability and death rate. So far the primary ways for human fall detection are image analysis, acoustic analysis and wearable analysis. Among all of these ways, wearable detection is a way that can not only meet the needs of fall detection system but also portable,with low energy as well as small in size.Therefore, this article mainly adopt wearable detection method to start the research.Firstly, the human activity patterns has been classified into ADL and falls; and the activity mode of fall experiment will be set and organized on this basis. By means of the moving posture data that collected through the acceleration sensor wearing in the corresponding part to analyze the acceleration data change of daily move and fall.Secondly, this paper has analyzed the typical wearable fall detection algorithm, and detected based on the threshhold value, putforward multipolar detection algorithm of convergence acceleration and posture angle due to the deficiency of traditional threshold value of fall detection algorithm. which not only promote the accuracy of detection but also can judge the exact human posture.The overall design of fall detection device has been made in this paper. AXDL345 triaxial accelerometer sensor and data acquisition module MSP430F149 microcontroller has been adopted as the data collection module and data processing module. Through the two ways of on-site alarm and active alarm, the wearer will send their falling data and location to rescuer, and the rescuer will take steps to rescue according to the data.Finally, though testing, the accuracy for detection of falls can be more than 97%. So the fall detection algorithm designed in this paper can achieve the goal setted in advance.