Research On Human Gesture Recognition Based On Motion Attitude Sensor MPU6050

Human gesture recognition technology,as a kind of human-computer interaction technology,is widely used in intelligent monitoring,robot,human-computer interaction,virtual reality,intelligent home,intelligent security,athlete auxiliary training and so on.With the rapid development of microelectronics technology,wearable devices based on motion attitude sensor have become the best method of human gesture recognition.At present,the human gesture recognition technology has been concerned by researchers at home and abroad,and the research on this aspect has made great progress.However,there are also problems such as inconvenient use of wearable devices,easy to be affected by the environment,difficulty in feature extraction,and insufficient identification accuracy.To solve these problems,this paper adopts the STM32 microcontroller combined with motion sensor MPU6050 triaxial acceleration,angular velocity and Angle data,through gestures to zero velocity testing calibration method to design a kind of low cost,no gyroscope drift human gesture recognition system,and from the data preprocessing,feature extraction and classification recognition algorithm of the series of studies in three aspects,the main work and innovations are as follows:(1)To solve the problem of inconvenient wearing of gesture recognition equipment,a human gesture data acquisition terminal based on STM32 microcontroller and motion attitude sensor is designed.(2)Aiming at the problem that the gyroscope sensor has zero deviation during continuous operation,which leads to larger errors,a gesture zero-speed detection and calibration method is designed,which can reduce the inherent noise data in the gesture recognition process and improve the reliability of the original data through calibration.(3)Aiming at the separation of noise data and gesture data in the original data,a continuous acceleration vector amplitude determination method is designed,which can effectively extract the interval of gesture data and eliminate the impact of data burr.(4)In view of the feature extraction problem of motion attitude sensor data,interpolation algorithm is introduced to normalize the length of original gesture data of the same type but different lengths.Principal component analysis(PCA)is introduced to reduce the dimension of data while retaining the original characteristics of data.(5)Aiming at the problem of insufficient accuracy of gesture classification and recognition,the threshold classification algorithm and linear discriminant classification algorithm are designed to classify and recognize different types of gestures.The human gesture classification and recognition system is verified by designing six kinds of gestures,which are flipping and waving.The actual verification data prove that the zero-speed detection and calibration method can effectively reduce noise.The gesture data can be accurately classified after feature extraction by the continuous acceleration vector amplitude determination method and dimensionality reduction by the principal component analysis method.The acquisition terminal equipment is low cost,simple and convenient to use,and has certain versatility and practicability.