Download presentation
Presentation is loading. Please wait.
Published byMolly Brown Modified over 9 years ago
1
Deconvolution in Reaction Kinetics Ernő Keszei Eötvös University Budapest, Hungary
2
mérendő görbe impulzus ( műszer válaszfüggvénye ) Measured signal Effect of convolution on kinetic signals Instrumental response function amplitude time (instantaneous) kinetic signal
3
For a continuous function : dt ' For discrete (measured) data points: imim olol smlsml What is convolution? “spread” “object” = “image”
5
For a continuous function : dt ' For discrete (measured) data points: imim olol smlsml What is deconvolution? “spread” “object” = “image”
6
Methods of deconvolution a priori knowledge of a kinetic model needed long computation time needed estimated kinetic parameters correlate with pulse parameters example: reconvolution simplicity short computation time needed example: Van Cittert’s method inverse filtering complicated computations long computation time needed example: Jansson’s method Bayes deconvolution Linear methods Nonlinear methods “Pseudo-deconvolution“ methods Direct deconvolution methods
7
Continuous Fourier transformation: Discrete Fourier transformation: Fourier transformation amplitude channel amplitude frequency
8
Inverse Fourier transformation yields the object function: Convolution in frequency space: I ( S ( · O ( Deconvolution in frequency space: O ( S (S ( I (I ( Inverse filtering using Fourier transforms ( “filtering” ) ( “inverse filtering” )
9
1. Fourier transformation of the measured signal Inverse filtering using Fourier transforms 2. Inverse filtering of the Fourier transform 3. Inverse Fourier transformation of the filtered result deconvolved signal
10
Inverse Fourier transformation : Deconvolution in frequency space: O ( S (S ( I (I ( Inverse filtering using Fourier transforms
11
Van Cittert (iterative) deconvolution Measured signal amplitude channel
12
Van Cittert (iterative) deconvolution convolved measured amplitude channel
13
Van Cittert (iterative) deconvolution measured convolved correction amplitude channel
14
Van Cittert (iterative) deconvolution measured convolved 1 st approximation of object function correction amplitude channel
15
Jansson deconvolution relaxation function amplitude channel
16
Jansson iteration: relaxation function
21
Bayes deconvolution o(x) (k) Probability theory based method ( Bayesian estimation ) previous object function measured signal o(x) (k+1) = new object function estimate s correction
22
Deconvolution via inverse filtering kinetic model function (instantaneous) channel amplitude
23
channel convolved (“measured”) signal (noise added) Instantaneous (modeled) signal Deconvolution via inverse filtering amplitude
24
convolved (measured) signal amplitude spectrum of the measured signal Deconvolution via inverse filtering channel amplitude
25
Amplitude spectrum of the deconvolved signal (NO filtering) Not applicable, due to high frequency noise (below) Deconvolution via inverse filtering channel amplitude convolved (measured) signal
26
Deconvolution via inverse filtering channel amplitude convolved (measured) signal Amplitude spectrum of the deconvolved signal (NO filtering)
27
deconvolved Deconvolution via inverse filtering channel amplitude Not applicable, due to high frequency noise (below) Modification: Replace high frequency part with exponential decay Or: use filter to smooth it amplitude spectrum of the deconvolved signal after filtering
28
amplitude original model curve Deconvolution via inverse filtering channel Not applicable, due to high frequency noise (below) Modification: Replace high frequency part with exponential decay Or: use filter to smooth it deconvolved amplitude spectrum of the deconvolved signal after filtering
29
special extrapolation of the measured signal prior to inverse filtering deconvolved signal Deconvolution via inverse filtering channel amplitude
30
deconvolved signal fitted model function special extrapolation of the measured signal prior to inverse filtering Deconvolution via inverse filtering channel amplitude
31
Bayes iteration 4 Bayes deconvolution results iteration step 4.4. deconvolved convolved amplitude channel
32
Bayes iteration 16 Bayes deconvolution results iteration step 16. deconvolved convolved amplitude channel
33
Bayes deconvolution results iteration step 128. deconvolved convolved Bayes iteration 128 amplitude channel
34
Bayes deconvolution results iteration step 512. deconvolved convolved Bayes iteration 512 amplitude channel
35
Bayes deconvolution results iteration step 1883. deconvolved original (model) function Bayes iteration 1883 amplitude channel
36
Comparison of deconvolution methods adaptáció után inverse filtering Jansson Bayes reconvolution inverse filtering Jansson Bayes reconvolution inverse filtering Jansson Bayes reconvolution inverse filtering Jansson Bayes reconvolution deviation / % adaptation only 1 st amplitude 2 nd amplitude risetime decay time
37
Comparison of methods inverse filtering Jansson Bayes reconvolution inverse filtering Jansson Bayes reconvolution inverse filtering Jansson Bayes reconvolution inverse filtering Jansson Bayes reconvolution deviation / % továbbfejlesztés után after modifications 1 st amplitude 2 nd amplitude risetime decay time
38
Summary What was I not talking about? technical details of practical applicability (the devil is hiding in the details) optimization of noise filtering applications in femtochemistry convolution in reaction kinetics applicable deconvolution methods adaptation to kinetic signals modification of standard deconvolution techniques inference capabilities of adapted, modified methods What was I talking about?
39
Your questions... Questions
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.