Provider: ingentaconnect
Database: ingentaconnect
Content: application/x-research-info-systems
TY - ABST
AU - Kawata, Satoshi
AU - Minami, Shigeo
TI - Adaptive Smoothing of Spectroscopic Data by a Linear Mean-Square Estimation
JO - Applied Spectroscopy
PY - 1984-01-01T00:00:00///
VL - 38
IS - 1
SP - 49
EP - 58
KW - Smoothing algorithm
KW - Computer, applications
KW - Data processing
N2 - An adaptive smoothing method based on a least mean-square estimation is developed for noise filtering of spectroscopic data. The algorithm of this method is nonrecursive and shift-varying with the local statistics of data. The mean and the variance of the observed spectrum at an individual
sampled point are calculated point by point from its local mean and variance. By this method, in the resultant spectrum, the signal-to-noise ratio is maximized at any local section of the entire spectrum. Experimental results for the absorption spectrum of ammonia gas demonstrate that this
method distorts less amount of signal components than the conventional smoothing method based on the polynomial curve-fitting and suppresses noise components satisfactorily. The computation time of this algorithm is rather shorter than that of the convolution algorithm with seven weighting
coefficients. The *a priori* information for the estimation of the signal by this method are: the variance of noise, which can be attainable in the experiment; and the window function which gives the local statistics. The investigation of various types of window functions shows that the
selection of the window function does not directly affect the performance of adaptive smoothing.
UR - http://www.ingentaconnect.com/content/sas/sas/1984/00000038/00000001/art00012
M3 - doi:10.1366/0003702844554305
UR - http://dx.doi.org/10.1366/0003702844554305
ER -