nspabav.m

一种新的时频分析方法的matlab源程序。

function [h,xs,w] = nspabav(data,nyy,minw,maxw,t0,t1,m)
%
%    [h,xs,w] = nspabav(data,nyy,minw,maxw,t0,t1,m):
%
%    Function to generate a smoothed HHT spectrum of data(n,k)  
%    in time-frequency space, where  
%    n specifies the length of time series, and 
%    k is the number of IMF components.
%    The frequency-axis range is prefixed.
%
%    Input-
%	data	- 2-D matrix data(n,k) of IMF components
%	nyy	- the frequency resolution
%	minw	- the minimum frequency
%	maxw	- the maximum frequency
%	t0	- the start time
%	t1	- the end time
%       m	- time scale averaging number
%    Output-
%	h	- 2-D matrix of the HHT spectrum, where
%		  1st dimension specifies the number of frequencies,
%		  2nd dimension specifies the number of time values
%	xs	- vector that specifies the time-axis values
%	w	- vector that specifies the frequency-axis values
%
%    Z. Shen (JHU)		July 2, 1995 Initial
%    J. Marshak (NASA GSFC)	Jan. 28, 2004 Edited
%
%    Notes-
%    MATLAB library function 'hilbert()' is used to calculate the
%    Hilbert transform.
%    Example, [h,xs,w] = nspabav(lod78_p',200,0,0.12,1,3224,3),
%    here m=3 means the time-axis will be 3 to 1.
%    Functions 'contour()' or img() can be used to view the spectrum,
%    for example contour(xs,w,h) or img(xs,w,h).
%
%    Temporary remarks-
%    Changed the function name,
%    was  'nspmab()' for the code named as 'nspmab.m'.
%    'nspabav.m' is similar to 'nspab.m': the latter has m=3 and 
%     does 3:1 averaging of time scale.
%     Question : Do we really need this option?
%

%----- Get dimensions (number of time points and components)
[npt,knb] = size(data);

%----- Get time interval
dt=(t1-t0)/(npt-1);

%----- Apply Hilbert Transform
data=hilbert(data);
a=abs(data);
omg=abs(diff(unwrap(angle(data))))/(2*pi*dt);

%----- Smooth amplitude and frequency
filtr=fir1(8,.1);
for i=1:knb
   a(:,i)=filtfilt(filtr,1,a(:,i));
   omg(:,i)=filtfilt(filtr,1,omg(:,i));
end
for i=1:knb
   a(:,i)=filtfilt(filtr,1,a(:,i));
   omg(:,i)=filtfilt(filtr,1,omg(:,i));
end

%----- Limit frequency and amplitude
for i=1:knb
   for i1=1:npt-1
      if omg(i1,i) >=maxw,
         omg(i1,i)=maxw;
	 a(i1,i)=0;
      elseif omg(i1,i)<=minw,
         omg(i1,i)=minw;
	 a(i1,i)=0;
      else
      end
   end
end
clear filtr data

%----- Get local frequency
dw=maxw - minw;
wmx=maxw;
wmn=minw;

%----- Construct the ploting matrix
clear p;
h1=zeros(npt-1,nyy+1);
p=round(nyy*(omg-wmn)/dw)+1;
for j1=1:npt-1
   for i1=1:knb
      ii1=p(j1,i1);	
      h1(j1,ii1)=h1(j1,ii1)+a(j1,i1);
   end
end

%----- Do m-point to 1-point averaging
[nx,ny]=size(h1);
n1=fix(nx/m);
h=zeros(n1,ny);
for i1=1:n1
   for j1=1:m
      h(i1,:)=h(i1,:)+h1(m*i1-(j1-1),:);
   end
   h(i1,:)=h(i1,:)/m;   
end
clear h1;

%----- Do m-points smoothing in x-direction
fltr=1./m*ones(m,1);
for j1=1:ny
   h(:,j1)=filtfilt(fltr,1,h(:,j1));
end
clear fltr;

%----- Define the results
w=linspace(wmn,wmx,ny-1)';
xs=linspace(t0,t1,n1)';
h=flipud(rot90(h));
h=h(1:ny-1,:);