Premium
2D wavelet analysis and compression of on‐line industrial process data
Author(s) -
Trygg Johan,
KettanehWold Nouna,
Wallbäcks Lars
Publication year - 2001
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/cem.681
Subject(s) - wavelet , wavelet transform , discrete wavelet transform , pattern recognition (psychology) , outlier , computer science , artificial intelligence , residual , anomaly detection , principal component analysis , data compression , second generation wavelet transform , mathematics , stationary wavelet transform , algorithm
In recent years the wavelet transform (WT) has interested a large number of scientists from many different fields. Pattern recognition, signal processing, signal compression, process monitoring and control, and image analysis are some areas where wavelets have shown promising results. In this paper, 2D wavelet analysis and compression of near‐infrared spectra for on‐line monitoring of wood chips is reviewed. We introduce a new parameter for outlier detection, distance to model in wavelet space (DModW), which is analogous to the residual parameter (DModX) used in principal component analysis (PCA) and partial least squares analysis (PLS). Additionally, we describe the wavelet power spectrum (WPS), the wavelet analogue of the power spectrum. The WPS gives an overview of the time–frequency content in a signal. In the example given, wavelets improved the detection of spectral shift and compressed data 1000‐fold without degrading the quality of the 2D wavelet‐compressed PCA model. The example concerned an industrial process‐monitoring situation where near‐infrared spectra are measured on‐line on top of a conveyer belt filled with wood chips at a Swedish pulp plant. Copyright © 2001 John Wiley & Sons, Ltd.