S-method with NUFFT



In [1]:

    
using PyPlot
import DSP



In [2]:

    
include("../juwvid.jl")









    Out[2]:





juwvid



In [3]:

    
# multicomponent data (modified Example 6.2.2)
nsample=200
t,x=sampledata.genmultifm622x(nsample);
PyPlot.plot(t,x)









    












    Out[3]:





1-element Array{PyCall.PyObject,1}:
 PyObject <matplotlib.lines.Line2D object at 0x7ff8db6b35f8>



In [4]:

    
# Wigner Ville suffers the cross talk
z=DSP.Util.hilbert(x);
tfr=cohenclass.tfrwv(z);
PyPlot.imshow(abs.(tfr[end:-1:1,:]),aspect=0.5)









    



Single Wigner Ville






    












    



Use fft.






    Out[4]:





PyObject <matplotlib.image.AxesImage object at 0x7ff8db640c18>



In [5]:

    
# STFT does not, but it's a poor resolution
tfrstfta=stft.tfrstft(x);
PyPlot.imshow(abs.(tfrstfta[100:-1:1,:]))









    












    



Use fft.






    Out[5]:





PyObject <matplotlib.image.AxesImage object at 0x7ff8da8a4a20>



In [6]:

    
# The results of the STFT w/ NUFFT shows that the poor resolution is not due to the sampling rate
fin=collect(linspace(10,75,nsample*16));
tfrstft=stft.tfrstft(x,NaN,NaN,fin,NaN,NaN,4);
PyPlot.imshow(abs.(tfrstft[end:-1:1,:]),aspect=0.03)









    












    



Use nufft.






    Out[6]:





PyObject <matplotlib.image.AxesImage object at 0x7ff8da882160>



In [7]:

    
# The S-method w/o NuFFT suppresses the cross talk. 
sm=smethod.tfrsm(x,NaN,5,NaN,2)
PyPlot.imshow(Real.(sm[100:-1:1,:]))









    



Single S-method






    












    



Use fft.






    Out[7]:





PyObject <matplotlib.image.AxesImage object at 0x7ff8da5cbf98>



In [8]:

    
# The S-method w/ NuFFT improves the sampling rate.
fin=collect(linspace(10,75,nsample*8));
sm=smethod.tfrsm(x,NaN,100,fin,2)
PyPlot.imshow(Real.(sm[end:-1:1,:]),aspect=0.06)









    



Single S-method






    












    



Use nufft.






    Out[8]:





PyObject <matplotlib.image.AxesImage object at 0x7ff8da5aa4e0>



In [ ]: