Anti-outlier noise reduction method based on Kalman filtering and least square fitting

Abstract

The invention discloses an anti-outlier noise reduction method based on Kalman filtering and least square fitting, and relates to signal noise reduction and outlier elimination in the field of signal processing. The method comprises the following steps: filtering a signal through employing a Kalman filtering algorithm, judging whether an observation value in the signal is an outlier or not through employing the Lett criterion in the filtering process, if the observed value is a normal value, and adopting the observed value to correct the Kalman predicted value at the moment so as to obtain a final Kalman filtering result; and if the observed value is an outlier, fitting an estimated value at the moment by adopting a least square method, and correcting a Kalman predicted value by utilizing the fitted estimated value, and in the correction process, and designing a regulatory factor which can gradually reduce the correction effect of the fitted estimated value on the predicted value when the outlier continuously appears so as to improve the filtering precision when the outliers continuously appear. The noise influence in the signal can be effectively reduced while the outlier interference is eliminated, and the identifiability of the signal is improved.

Description

technical field [0001] The invention belongs to the field of signal processing, in particular to an outlier noise reduction method based on Kalman filtering and least square fitting. Background technique [0002] In engineering testing, due to the interference of the measurement equipment itself, random errors, and various environmental factors, the measurement data is noisy, and some measurement data will also contain data that seriously deviates from the true value of the target. Usually, these seriously deviate from the true value of the target. The data of the outlier is called "outlier". According to whether it is continuous, the outlier can be divided into isolated outlier and continuous outlier. The form of isolated outlier is isolated point, while the continuous outlier appears in patches It is called speckle-type outlier. When denoising the measurement data, the outlier will bring a large error to the noise reduction result, and even seriously distort the signal. ...

AI Technical Summary

Problems solved by technology

[0002] In engineering testing, due to the interference of the measurement equipment itself, random errors, and various environmental factors, the measurement data is noisy, and some measurement data will also contain data that seriously deviates from the true value of the target. Usually, these seriously deviate from the true value of the target. The data of the outlier is called "outlier". According to whether it is continuous, the outlier can be divided into isolated outlier and continuous outlier. The form of isolated outlier is isolated point, while the continuous outlier appears in patches It is called speckle-type outlier. When denoising the measurement data, the outlier will bring a large error to the noise reduction result, and even seriously distort the signal.

Images