Design And Implementation Of Bone Point-based Human Motion Recognition System

The current era is the era of rapid development of artificial intelligence.Human-computer interaction technology has entered the white-hot stage with the development of artificial intelligence technology.Motion capture technology is an important branch of human and computer natural,multi-modal interaction technology.In recent years,it has developed rapidly.In this paper,an inertial sensor is designed and implemented to realize a system based on skeleton point attitude information,which can recognize human motion.The system collects the sensor data located at the main skeleton point of the human body,transmits it to the upper computer through multi-Bluetooth pairing transmission,and uses the support vector machine to classify the human motion to realize the recognition of the general human motion.The system is light and convenient,easy to use,adaptable to the environment,and low in cost.In this paper,we first investigate the related work of human body motion analysis and recognition technology by domestic and foreign research scholars,then introduce the relevant knowledge and basic theory involved in the design process of this system,and design and implement a human body based on this The motion recognition system of skeletal point attitude information is studied in detail for the design requirements and characteristics of different functional modules.Firstly,according to the characteristics of multi-node,real-time,dynamic and easy identification of the system,based on the attitude sensor solution theory,calibration technology and sensor networking technology,the hardware experimental platform based on MPU6050 and Bluetooth 4.0 is determined.Secondly,according to the characteristics of human body movement,under the premise of low cost,combined with the experimental environment,a reasonable data acquisition scheme is designed,and the motion information of the human body is constructed by using the scheme to collect the motion information of the human skeleton.Finally,the collected skeleton point attitude information is analyzed,and the machine learning classification algorithm such as support vector machine is used to classify the motion pose,and the different classification algorithms are compared.The whole system is based on modularization and light weight,organically integrates various key technologies,and designs a human motion recognition system based on bone point attitude information.The experimental results show that the attitude data of the collection node of the system can be collected by means of multiple Bluetooth direct connection,which can ensure real-time performance.The system can collect the human motion data sets formed by the motion information of different testers,and can be classified by various classification methods,and the different motion recognition accuracy rates are high,which can meet the design system design requirements.In the test based on 1762 data,the support vector machine is used for classification training,and the accuracy of human motion recognition reaches 88%,among which the static action classification accuracy can reach 90%,and the dynamic action classification classification accuracy can reach 80%.The system is light and simple,simple to use,and dynamic follow-up number,which overcomes the spatial limitations of optical motion detection methods and wired acquisition methods,and can be applied to sports exercises,human-computer interaction,medical diagnosis and game production.