传感器输出时间序列的实时小波滤波方法

为了满足不同测控系统中传感器输出时间序列的实时滤波要求，针对小波变换不能递推进行的弱点，通过分析小波实时滤波算法，研究了基于Mallat快速离散小波变换的3种实时滤波方法：冗余采样频率法、有滞后的浮动数据窗法和对称添加的浮动数据窗法，给出了3种方法的实现过程及适应范围，分析了影响小波实时滤波中的几个关键因素。仿真实验结果表明，本文讨论的3种实时滤波方法，运行时间短，滤波性能与离线滤波相接近，可适用于一般流程工业的大部分参数变化较慢的测控系统。

Real time wavelet filtering methods for sensor output time series

In order to meet the needs of real time wavelet filtering for sensor output time series in various control systems,it is necessary to overcome the problem that the wavelet transform cannot be treated by iterative computations.We report an investigation of three real time filtering methods,namely the redundant sampling method,the lagged float data window method and the symmetrically added float data window method,based on Mallat's Fast Discrete Wavelet and analysis of the real time wavelet filtering algorithm.The principles and range of applicability of the three filtering methods are summarized,and the key factors affecting real time wavelet filtering are identified.The results of simulation experiments show that these three filtering methods are characterized by short runtimes and filtering performance close to off-line filtering.It is shown that the methods may be used in measurment and control systems in which the process parameters vary relatively slowly.