基于无迹卡尔曼滤波和小波分析的IMU传感器去噪技术研究  

IMU sensor denoising based on unscented Kalman filter and wavelet analysis

作　　者：阳兆哲 李跃忠[1] 吴光文[1] YANG Zhaozhe;LI Yuezhong;WU Guangwen(School of Mechanical and Electronic Engineering,East China University of Technology,Nanchang 330032,China)

机构地区：[1]东华理工大学机械与电子工程学院,江西南昌330032

出　　处：《现代电子技术》2024年第5期53-59,共7页Modern Electronics Technique

基　　金：江西省科技厅重点研发一般项目(20212BBE53033)。

摘　　要：获得精确的姿态信息是跌倒检测的关键。文中在姿态角解算问题中提出一种基于无迹卡尔曼滤波和小波滤波的改进方法,通过Savitzky-Golay滤波器和小波滤波融合算法对加速度计以及陀螺仪数据进行降噪处理,利用降噪后的加速度数据对陀螺仪数据进行PI积分补偿,将补偿后的陀螺仪数据进行Mahony解算,其结果作为无迹卡尔曼滤波的状态信息;其次通过加速度值解算,将其结果作为无迹卡尔曼滤波的量测信息实现姿态解算。实验表明,在静态条件下,相对于常见的扩展卡尔曼滤波融合切比雪夫滤波算法,该方法使IMU传感器原始加速度计精度提高了83.3%,姿态角标准差平均减少了0.00193,能够有效地减少随机噪声。零点漂移、高斯噪声对IMU传感器姿态角信号的影响,使跌倒检测系统在复杂的环境条件下具有较高的精度以及稳定性。Getting accurate attitude information is the key to fall detection.In this paper,an improved method based on unscented Kalman filter(UKF)and wavelet filtering is proposed for attitude angle estimation.The data of accelerometer and gyroscope are denoised by Savitzky⁃Golay filter and wavelet filter fusion algorithm.The denoised acceleration data are used for PI integral compensation for gyro data.The compensated gyro data are subjected to the Mahony solution,and the results of solution is used as the state information of the UKF.And then,the acceleration value is used to solve the problem,and the results are used as UKF measurements for attitude determination.Experiments show that the accuracy of the original accelerometer of the IMU(inertial measurement unit)sensor is improved by 83.3%in comparison with the conventional extended Kalman filter fusion Chebyshev filtering algorithm under static conditions,and the mean attitude standard deviation is reduced by 0.00193,so the improved method can effectively reduce the random noise.Zero drift and Gaussian noise have influence on the attitude angle signal of the IMU sensor,which makes the fall detection system more accurate and stable under complex environmental conditions.

关 键 词：跌倒检测 小波滤波 Savitzky-Golay滤波器 无迹卡尔曼滤波 IMU传感器 姿态角 

分 类 号：TN967.2-34[电子电信—信号与信息处理]