基于稀疏表示的惯性传感器信号实时滤波方法  

Real-time filtering method based on sparse representation for inertial sensor signals

作　　者：蒋行国[1] 张龙[1] 许金海[1] 

机构地区：[1]桂林电子科技大学信息与通信学院,广西桂林541004

出　　处：《计算机应用研究》2015年第5期1480-1482,共3页Application Research of Computers

摘　　要：针对惯性传感器信号处理的特点,提出了基于稀疏表示的信号滤波处理系统模型和方法。通过K-SVD算法对信号学习训练获得字典,为了减少计算量,满足实时性,尽量降低字典的大小,仿真结果表明,在满足一定精度的条件下,字典的大小最小为3×10。在该字典下对信号进行稀疏表示和重构,改变信号的输入方式,可以实现信号的实时滤波。仿真结果表明提出的滤波方法能有效地消除噪声,改善输出信号精度,可以提高信噪比最大为4.5 d B。该滤波方法与传统的滤波方法相比有较大的优势,为惯性传感器信号处理提供了一种新的方法。According to the characteristics of inertial sensor signal processing,this paper proposed the signal processing system and method based on sparse representation. Through the K-SVD algorithm of signal learning training to obtain the dictionary,and in order to reduce the amount of computation to meet the real-time,as far as possible to reduce the size of the dictionary,the simulation results show that,under certain precision condition,the minimum size of the dictionary is 3 × 10. Then,in the dictionary under the sparse representation and reconstruction of signals,while changing the input signal,it could achieve signal real-time filtering. The simulation results show that the proposed method can effectively eliminate the noise,improve the accuracy of output signals,and improve SNR up to 4. 5 d B. The filtering method has more advantages than the traditional method,and provides a new method for inertial sensor signal processing.

关 键 词：惯性传感器 稀疏表示 字典 实时滤波 

分 类 号：TN911.7[电子电信—通信与信息系统]