4/1/13
Signal Reconstruction from Continuous Wavelet Transform Coefficients - MATLAB & Simulink Example - MathWorks India
Signal Reconstruction from Continuous Wavelet Transform Coefficients
The continuous w avelet transform (CWT) allow s you to analyze the temporal evolution of the frequency content of a signal or time series. The CWT can be used to detect time-localized events and scale-localized components. When the Fourier transform of a w avelet satisfies certain properties, it is possible to invert the CWT by integrating only over scales. Accordingly, Fourier transform based CWT and inverse CWT algorithms w ith select w avelets enable you to reconstruct a time and scale-localized approximation to a signal. Step 1: Detection of System Anom aly Using the CWT The data analyzed in this example w ere generously provided by the oil company Total. These data are also used in the w avelet coherence example. In that example, the w avelet coherence betw een signals recorded at different sensor locations enables you to detect a system anomaly indicated by the appearance of an oscillatory component. Since it is possible to detect this component by examining data recorded at an individual sensor, w e restrict our attention here to the signal at sensor 1. Analyzing the time series recorded at sensor 1 w ith the CWT reveals the quasi-periodic signal indicative of a system anomaly.
%D e f i n et h es i g n a lt oa n a l y z e . c l e a ra l l ; l o a ds e n s o r 1 ; l o n g =4 0 0 0 ; f i r s t=5 0 0 0 ; l a s t =f i r s t+l o n g 1 ; i n d i c e s=( f i r s t : l a s t ) ; d t=1 ; s i g n a l=s e n s o r 1 ( i n d i c e s ) ; s 1 { 1 }=s i g n a l ; s 1 { 2 }=d t ;
Obtain the CWT over scales 1 to 50 in 0.1 increments using the analytic Morlet w avelet, ' m o r l ' . With the bandw idth and center frequency settings used in the w avelet coherence example, the oscillatory behavior in the sensor data w as observed at approximately scale 15. This scale translates approximately to scale 6 w ith the analytic Morlet w avelet used here. Define the parameters and perform the CWT analysis.
s c a l e s=1 : 0 . 1 : 5 0 ; w n a m e =' m o r l ' ; p a r W A V =6 ; ={ w n a m e , p a r } ;
c w t _ s 1 _ l i n= c w t f t ( s 1 , ' s c a l e s ' , s c a l e s , ' w a v e l e t ' , W A V , ' p l o t ' ) ;
www.mathworks.in/help/wavelet/examples/signal-reconstruction-from-continuous-wavelet-transform-coefficients.html
1/8
4/1/13
Signal Reconstruction from Continuous Wavelet Transform Coefficients - MATLAB & Simulink Example - MathWorks India
The figure show s the CWT decomposition. The middle figure panel in the left column contains a plot of the CWT moduli revealing strong components around scale 6 over approximately the first 2000 points. Another w ay to compute the CWT coefficients is to use the logarithmically-spaced scales provided by the default values.
c w t _ s 1 _ p o w=c w t f t ( s 1 , ' p l o t ' ) ;
The output c w t _ s 1 _ p o wis a structure array w hich contains six fields: the CWT coefficients (c f s ), the vector of scales, the angular frequencies used in the Fourier transform in radians/sample ( o m e g a ), the mean of the signal ( m e a n S I G ), the sampling period ( d t ), and the w avelet ( w a v ). You can use i c w t f tto reconstruct the original signal based on the CWT coefficients.
www.mathworks.in/help/wavelet/examples/signal-reconstruction-from-continuous-wavelet-transform-coefficients.html 2/8
4/1/13
Signal Reconstruction from Continuous Wavelet Transform Coefficients - MATLAB & Simulink Example - MathWorks India
r e c _ s 1 _ p o w=i c w t f t ( c w t _ s 1 _ p o w ) ;
Utilizing a subset of the CWT coefficients, you can reconstruct an approximation to the original signal. The next example illustrates this for the oscillatory component in the sensor 1 data. As a preliminary step, determine the scale containing the greatest percentage of the signal energy.
%C o m p u t et h ee n e r g yd i s t r i b u t i o no v e rs c a l e s . c f s=c w t _ s 1 _ l i n . c f s ; e n e r g y=s u m ( a b s ( c f s ) , 2 ) ; p e r c e n t a g e=1 0 0 * e n e r g y / s u m ( e n e r g y ) ; %D e t e c tt h es c a l eo fg r e a t e s te n e r g y . [ m a x p e r c e n t , m a x S c a l e I D X ]=m a x ( p e r c e n t a g e ) f i g u r e ; p l o t ( p e r c e n t a g e , ' . r ' ) ;h o l do n ; p l o t ( m a x S c a l e I D X , m a x p e r c e n t , ' . k ' , ' M a r k e r s i z e ' , 2 0 ) x l a b e l ( ' I n d i c e so fS c a l e s ' ) y l a b e l ( ' P e r c e n t a g eo fe n e r g y ' ) ; a x i st i g h t g r i d %T h es c a l eo fg r e a t e s te n e r g yi sg i v e nb y : s c a M a x E n e r=s c a l e s ( m a x S c a l e I D X )
m a x p e r c e n t= 0 . 3 5 8 9
m a x S c a l e I D X= 4 1
s c a M a x E n e r= 5
www.mathworks.in/help/wavelet/examples/signal-reconstruction-from-continuous-wavelet-transform-coefficients.html
3/8
4/1/13
Signal Reconstruction from Continuous Wavelet Transform Coefficients - MATLAB & Simulink Example - MathWorks India
Step 2: Reconstruction of System Anom aly Signature in the Tim e Dom ain An interesting use of the analysis-reconstruction process involves modifying the w avelet coefficients betw een analysis and reconstruction. This is routinely done for denoising or compression applications using the discrete w avelet transform. Because the Fourier transform based CWT enables us to efficiently implement an inverse transform, it is possible to extend this selective reconstruction to the CWT. Using this technique, w e can obtain a clearer picture of the system anomaly.
c w t _ a n o m a l y=c w t _ s 1 _ p o w ; %F i n dt h ei n d e xo fl o g a r i t h m i cs c a l ed e t e c t i n gt h e a n o m a l y . [ v a l M i n , a n o m a l y _ i n d e x _ s c a l e s ]=m i n ( a b s ( c w t _ s 1 _ p o w . s c a l e s s c a M a x E n e r ) )
v a l M i n= 0 . 5 1 1 7
a n o m a l y _ i n d e x _ s c a l e s= 4
a n o m a l y _ c f s=c w t _ s 1 _ p o w . c f s ( a n o m a l y _ i n d e x _ s c a l e s , : ) ; n e w C F S=z e r o s ( s i z e ( c w t _ s 1 _ p o w . c f s ) ) ; n e w C F S ( a n o m a l y _ i n d e x _ s c a l e s , : )=a n o m a l y _ c f s ; c w t _ a n o m a l y . c f s=n e w C F S ; %R e c o n s t r u c t i o nf r o mt h em o d i f i e ds t r u c t u r e . a n o m a l y=i c w t f t ( c w t _ a n o m a l y , ' p l o t ' , ' s i g n a l ' , s 1 ) ;
www.mathworks.in/help/wavelet/examples/signal-reconstruction-from-continuous-wavelet-transform-coefficients.html 4/8
4/1/13
Signal Reconstruction from Continuous Wavelet Transform Coefficients - MATLAB & Simulink Example - MathWorks India
In this example only the reconstructed signal is of interest. Plots of the CWT coefficients are not informative because only a single scale contains nonzero coefficients. To obtain a better view of the anomaly, you can perform a horizontal zoom for all axes of the figure.
a x=f i n d o b j ( g c f , ' t y p e ' , ' a x e s ' , ' t a g ' , ' ' ) ; s e t ( a x , ' X l i m ' , [ 2 5 05 0 0 ] ) ;
The selective reconstruction process may be done more directly using the IdxSc name-value pair in the i c w t f tfunction.
i c w t f t ( c w t _ a n o m a l y , ' p l o t ' , ' s i g n a l ' , s 1 , ' I d x S c ' , a n o m a l y _ i n d e x _ s c a l e s ) ;
www.mathworks.in/help/wavelet/examples/signal-reconstruction-from-continuous-wavelet-transform-coefficients.html 5/8
4/1/13
Signal Reconstruction from Continuous Wavelet Transform Coefficients - MATLAB & Simulink Example - MathWorks India
a x=f i n d o b j ( g c f , ' t y p e ' , ' a x e s ' , ' t a g ' , ' ' ) ; s e t ( a x , ' X l i m ' , [ 2 5 05 0 0 ] ) ;
Step 3: A Second Reconstruction of System Anom aly Signature Continuous w avelet analysis greatly facilitates the detection of the quasi-periodic component indicative of a system anomaly. In addition, the inverse CWT algorithm enables you to reconstruct a time and scale-localized approximation to the component based on selected coefficients. The Fourier transform based CWT and inverse CWT algorithms for logarithmic scales are described in Torrence & Compo, 1998. Sun, 2010 proposes an inversion formula for linearly-spaced scales. The algorithm w ith linearly-spaced scales is less efficient than its logarithmically-spaced counterpart because more scales are required for an accurate reconstruction. Nevertheless, this algorithm is experimentally useful and complements the algorithm w ith logarithmically-spaced scales. To end this example, w e illustrate the use of selective linearly-spaced scales for signal approximation.
%F i r s ts t e pf o rb u i l d i n gt h en e ws t r u c t u r ec o r r e s p o n d i n g t ot h ea n o m a l y . c w t _ a n o m a l y=c w t _ s 1 _ l i n ; %C h o o s eav e c t o ro fs c a l e sc e n t e r e do nt h em o s te n e r g e t i c s c a l e . d S c a l e=5 ; a n o m a l y _ i n d e x _ s c a l e s=( m a x S c a l e I D X d S c a l e : m a x S c a l e I D X + d S c a l e ) ; a n o m a l y _ c f s=c w t _ s 1 _ l i n . c f s ( a n o m a l y _ i n d e x _ s c a l e s , : ) ; n e w C F S=z e r o s ( s i z e ( c w t _ s 1 _ l i n . c f s ) ) ; n e w C F S ( a n o m a l y _ i n d e x _ s c a l e s , : )=a n o m a l y _ c f s ; c w t _ a n o m a l y . c f s=n e w C F S ; %R e c o n s t r u c t i o nf r o mt h em o d i f i e ds t r u c t u r e . a n o m a l y=i c w t l i n ( c w t _ a n o m a l y , ' p l o t ' ) ;
www.mathworks.in/help/wavelet/examples/signal-reconstruction-from-continuous-wavelet-transform-coefficients.html
6/8
4/1/13
Signal Reconstruction from Continuous Wavelet Transform Coefficients - MATLAB & Simulink Example - MathWorks India
To get a better view of the anomaly, you can perform a horizontal zoom for all axes of the figure.
a x=f i n d o b j ( g c f , ' t y p e ' , ' a x e s ' , ' t a g ' , ' ' ) ; s e t ( a x , ' X l i m ' , [ 2 5 05 0 0 ] ) ;
The selective reconstruction process may be done more directly using the IdxSc name-value pair in the i c w t l i nfunction.
i c w t l i n ( c w t _ a n o m a l y , ' p l o t ' , ' s i g n a l ' , s 1 , ' I d x S c ' , a n o m a l y _ i n d e x _ s c a l e s ) ; a x=f i n d o b j ( g c f , ' t y p e ' , ' a x e s ' , ' t a g ' , ' ' ) ; s e t ( a x , ' X l i m ' , [ 2 5 05 0 0 ] ) ;
www.mathworks.in/help/wavelet/examples/signal-reconstruction-from-continuous-wavelet-transform-coefficients.html
7/8
4/1/13
Signal Reconstruction from Continuous Wavelet Transform Coefficients - MATLAB & Simulink Example - MathWorks India
Sum m ary This example show s how to use continuous w avelet analysis to detect time and scale-localized components in the time-scale plane. Using Fourier transform based continuous w avelet transform (CWT) and inverse CWT algorithms, the example illustrates how to reconstruct a signal approximation based on scales identified in the analysis. References 1. Torrence, C. & Compo, G.C. A Practical Guide to Wavelet Analysis , Bull. Am. Meteorol. Soc., 79, 61-78, 1998. 2. Sun, W. Convergence of Morlet's Reconstruction Formula, Preprint, 2010.
Was this topic helpful?
Yes
No
Try MATLAB, Simulink, and Other Products
Get trial now
www.mathworks.in/help/wavelet/examples/signal-reconstruction-from-continuous-wavelet-transform-coefficients.html
8/8
