Research On Calibration Model Of Accelerometer Based On Improved Levenberg-Marquardt Algorithm

With the approach of the 2022 Beijing Winter Olympics,the popularity of ice and snow sports is getting higher and higher.In order to better promote ice and snow sports and train more professionals,it is necessary to provide scientific and effective training methods.Wearable devices can fully track the position and posture information of personnel during sports training.Data analysis will not only help coaches to comprehensively and in-depth understand the motion characteristics of sports personnel,but also evaluate technical movements and give effective instructions.MEMS(micro electro mechanical system)accelerometer is one of the main components of wearable equipment.In order to ensure the accuracy of feedback information,the accelerometer needs to be calibrated before use.In this paper,the calibration model of MEMS accelerometer is studied for the sensor data calibration and fusion problem.The main work of the paper is as follows:First,according to the calibration principle of MEMS accelerometers in wearable devices,a mathematical model based on errors is established.Determine the feasible calibration method based on the use environment and accuracy requirements of the wearable device.Due to the noise pollution of the collected data,after analyzing the noise characteristics,it is determined that the median filter is used to denoise the data.The results show that the filtering method can effectively reduce the noise of the accelerometer.Since the mathematical model based on the error is a nonlinear least squares optimization problem,the common parameter identification algorithms are introduced and the limitations in solving the calibration model of MEMS accelerometer are pointed out.Secondly,in order to optimize the parameter identification of the accelerometer calibration model,an improved Levenberg-Marquardt algorithm is proposed.In the process of parameter identification,aiming at the problem that the original Levenberg-Marquardt algorithm iteratively solves the problem of slow decline of the optimal estimate and large amount of computation,by making full use of the calculation results of each iteration of the algorithm to set the step size factor,the number of iterations to obtain the optimal estimate is reduced,and the convergence of the algorithm is proved theoretically.Numerical experiments are carried out with the obtained open source data,and the comparison between the results of the original algorithm and the improved algorithm shows that the iteration times of the improved algorithm are reduced and the precision of the calibration parameters is improved.Finally,in view of the problem that the value deviates from the real value after calibration,the paper proposes to improve the calibration model of MEMS accelerometer by using the state information of the sensor.The open source data and the measured data of wearable devices studied in this paper were used to carry out numerical experiments.The experimental results all have the problem that the value of the MEMS accelerometer deviates from the true value after calibration.The improved calibration model is used to solve such problems,and the feasibility and effectiveness of the improved Levenberg-Marquardt algorithm are also verified.