In this paper, a new type of adaptive filtering algorithm, which can adapti
vely remove all kinds of noises from signals of analytical instruments unde
r a variety of complex conditions, is proposed. At present, the popular fil
tering algorithms which are widely applied to the data processing equipment
for analytical instrument are low-pass filter or band-pass filter. The fun
damental of those filters depends on the fact that the frequency characteri
stics of real signals are different from those of noises. These filtering a
lgorithms based on the different frequency distribution characteristics bet
ween signals and noises have an obvious defect, that is, users have to pres
et properly initial filter factors according to the width of peaks, which g
reatly influences the objectivity and veracity of computational results in
analytical procedures. In the light of the wavelet transform modulus maximu
m theory proposed by Mallat, the characteristics of wavelet transform modul
us maxima of real signals are distinctively different from those of noises
in the practical signals of analytical instruments, such as chromatography.
It is easy to identify them. Taking advantage of the different characteris
tics between real signals and noises on different scales in wavelet transfo
rmation domain, noises can be removed from the practical signals or analyti
cal instruments while avoiding to distort the real signals. The adaptive fi
ltering algorithm designed by this principle breaches the popular patterns
of current filtering algorithms, and radically improves the filtering effec
ts. A lot of tests using chromatography data prove that this algorithm has
a serial of virtues, such as no requirement on artificially presetting filt
er factors, excellent separation of signals and noises, holding the positio
n and height of peaks, and so on. Its performance in the robustness, adapta
bility and fidelity of peak completely satisfy the needs of signal processi
ng for analytical instruments.