Research On The Attitude Detection And Recognition Technology Of Aged People In Centralized Pension Environment

All countries in the world,especially the developing countries,there are faced with a common topic: the aging of the population.According to the specific conditions of our country,large population flows,city workers,and lead to the left behind elder and children population.Centralized pension for government management is conducive to the development of the society,so the centralized pension will become the inevitable trend of social development.In the centralized pension environment based on the attitude detection and recognition of daily behavior detection to the elderly has become an important research field,which includes three types of behavior detection: daily life behavior,behavior,fall down,especially the fall behavior of the elderly will cause harm.There are many incidents of sprain,fracture,and death every year because of a fall.There are mainly the elderly attitude detection and identification scheme has three kinds: image or video detection system,detection system based on the environment,based on the wearable inertial sensor detection system.Based on the in-depth study of the advantages and disadvantages of the development trend of each system,wearable detection system based on inertial sensor has a strong potential for development.The wearable inertial products will become a part of the future of Smart clothes.In order to solve the problem of attitude detection and recognition of centralized pension under the environment of the elderly,the inertial sensors MPU9250 and STM32 controller to realize the human data acquisition,there are kept by assistant of the collected data,the characteristics of the human body movement and the characteristics of daily data collection analysis,human behavior is divided into 3 categories,respectively: Non-Fall,Semi-Fall,Fall.The Non-Fall actions include walking,slow walking,running,standing or lying down.The data characteristics of these actions are quite obvious.Only using human acceleration value,we can recognize by using threshold algorithm.In order to make this part of data more practical,one calculate the steps and calculate the distance of movement.One can judge whether the old people get the proper exercise through the size of movement.The actions of Semi-Fall mainly include running,up and down stairs,bending,squatting,jumping and other actions.Since the drift of the acceleration data cannot be determined by extracting effective features,software compensation of the original data is required.For fall class behavior,compared with other scholars,were identified in more detail,the 9 axis sensor of Euler angles are solved by Mathony data fusion algorithm,using acceleration with pitch angle and roll angle of the body to determine whether the direction of falls and fall.Finally,through the analysis of the algorithm evaluation results showed that the change of Euler angle,recognition of human fall direction reached around 95%;the step algorithm and recognition accuracy was 94% who using the acceleration data;Semi-Fall action needs further optimization of the algorithm.