Based On MEMS Sensors Human Body Elderly Assistance And Rehabilitation Motion Gesture Recognition Research

Human motion capture and detection technology is widely used in rehabilitation medical and elderly assistance services.The performance of MEMS devices is constantly improving with the continuous development of MEMS technology,and its unique advantages such as small size,light weight,and low price are becoming more prominent and more and more important to researchers.MEMS inertial sensors can be used to measure human body motion-related information such as acceleration and angular velocity,and then through calculations,the change in attitude angle can be obtained.Subsequently,in order to classify the human motion state,we can use the method of pattern recognition to achieve the purpose of capturing and detecting human motion.Based on MEMS wearable sensors,this article mainly studies human motion detection and gesture recognition methods.This method can provide external sports information for some assistance agencies or rehabilitation agencies.First,design the human body posture data collection device(mainly including the sensor module,control module,wireless transmission module and power supply module),and the collected posture data is transmitted to the PC through the wireless transmission module for data calculation.Process and analyze the original motion data(including data preprocessing,posture calculation and feature extraction).The noise is suppressed by the fourth-order Butterworth filter,and then the sensor data is calibrated.Secondly,the posture calculation is performed.First,the posture information is solved by the quaternion-based posture solution method,and then the Pica algorithm is used to solve the quaternion differential equation,and finally the real-time posture information is updated.The calculated posture information is transformed into the abdomen coordinate system through coordinate conversion to form a standard matrix representing the overall posture of the human body.Calculate the mean,variance,range,maximum,minimum,interquartile range and skewness features of the standard matrix,and evaluate the availability of features through the distribution of each posture feature.The final human posture recognition algorithm constructed is to divide the data into two categories of static posture data and motion posture data through the binary classifier of static posture and motion posture.The static posture data is classified and recognized in the binary decision tree,and the motion posture data is in the binary decision tree.The classification and recognition in the support vector machine,through the construction of a multi-layer recognition algorithm,completed the classification and recognition process of seven poses.Experiments show that when the system is turned on for 3 seconds,the recognition rate of the movement posture is 40.8%,which is relatively low.However,as the running time of the system increases,the amount of collected data increases.After the system is powered on for 11 seconds,the average recognition rate of static posture is over 99.9%,and the average recognition rate of motion posture is over 93.2%.The eigenvalues extracted based on the standard matrix can realize the classification and recognition of seven common human postures in rehabilitation institutions for the elderly after undergoing secondary classification.