Research On The Learning State Recognition Method Based On MARG Sensor

With the deep integration of information technology and education teaching,the way of learning is strongly impacted and online learning has become an important learning trend in the future.Therefore,it is very important to find out the learning rules and learning problems of learners in time in this unsupervised environment.In this context,based on theories of inertial measurement and pattern recognition,this thesis deeply investigates the key technologies of head movement attitude detection and pattern recognition under online learning state.The Magnetic,Angular Rate and Gravity(MARG)sensors-based learning state detection and recognition system is designed and implemented.This system can detect the attitude of head and recognize the pattern of head motion in real time.The main research works of the thesis are as follows:(1)Owing to the disadvantages of low precision and poor stability of the traditional attitude calculation algorithm,this thesis proposes a novel head attitude algorithm based on the information fusion of improved complementary filtering and Extended Kalman Filter.We firstly rely on the complementary filtering combined with PI algorithm to estimate the gyroscope drift error,and then realize the attitude estimation by using the Extended Kalman Filter.By comparison the performance of the conventional single Extended Kalman Filter and the proposed algorithm in filtering,the results demonstrate that the proposed algorithm can constraint drift and noise of gyroscope and accelerometer,and consequently improve the stability and precision of the head attitude algorithm.(2)The thesis proposes a novel method of learning state recognition which divides learning state pattern into three levels:attitude,movement and state.Based on the Support Vector Machine,the learning attitude recognition algorithm is designed to realize the recognition of seven basic postures:left-turn head,right-turn head,head-down,head-up,left-tilt head,right-tilt head and head-up.The learning state recognition algorithm is designed by using BP neural network,which realizes the recognition of four kinds of basic movements:nodding,shaking,turning and normal.On this basis,a state recognition algorithm based on threshold method is proposed to recognize the fatigue state and the inattention state in the online learning scenarios.(3)With CC1310 as the microprocessor and MPU9250 as the 9-axis sensor,the hardware system is designed.Then the head attitude detection experiment and the learning state recognition experiment are carried out.The results has shown that the improved attitude calculation algorithm can detect the head posture accurately.The classifier based on learning attitude recognition algorithm has a better effect on attitude recognition,and the average correct recognition rate reaches up to 98.1053%.The classifier based on learning movement recognition algorithm has a better effect on movement recognition,and the average correct recognition rate reaches up to 92.2200%The classifier based on learning state recognition algorithm has a better effect on the recognition of fatigue state and inattention state,and the average recognition rate reaches up to 90.8000%and 92%.The learning state detection and recognition system designed in this thesis realizes the detection and recognition of the learner's learning state,which is of great significance to standardize the learner's learning behavior and improve the learner's learning efficiency.