Noise level estimation using haar wavelet packet trees for sensor robust outlier detection

  • Authors:

  • Paolo Mercorelli;Alexander Frick

  • Affiliations:

  • Dep. of Vehicles, Production and Process Engineering, University of Applied Sciences Wolfsburg, Wolfsburg, Germany;ABB Utilities GmbH UTD/PAT2, Mannheim, Germany

  • Venue:
  • ICCSA'06 Proceedings of the 6th international conference on Computational Science and Its Applications - Volume Part I
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

The paper is related to the on-line noise variance estimation. In practical use, it is important to estimate the noise level from the data rather than to assume that the noise level is known. The paper presented a free thresholding method related to the on-line peak noise variance estimation even for signal with small S/N ratio. The basic idea is to characterize the noise like an incoherent part of the measured signal. This is performed through the wavelet tree by choosing the subspaces where the median value of the wavelet components has minimum. The paper provides to show nice general properties of the wavelet packets on which the proposed procedure is based. The developed algorithm is totally general even though is applied by using Haar wavelet packets and it is present in some industrial software platforms to detect sensor outliers. More, it is currently integrated in the inferential modeling platform of the Advanced Control and Simulation Solution Responsible Unit within ABB’s industry division.