Mercurial > hg > octave-lyh
view scripts/general/interpft.m @ 8920:eb63fbe60fab
update copyright notices
author | John W. Eaton <jwe@octave.org> |
---|---|
date | Sat, 07 Mar 2009 10:41:27 -0500 |
parents | e3041433a57e |
children | 1bf0ce0930be |
line wrap: on
line source
## Copyright (C) 2001, 2006, 2007, 2008, 2009 Paul Kienzle ## ## This file is part of Octave. ## ## Octave is free software; you can redistribute it and/or modify it ## under the terms of the GNU General Public License as published by ## the Free Software Foundation; either version 3 of the License, or (at ## your option) any later version. ## ## Octave is distributed in the hope that it will be useful, but ## WITHOUT ANY WARRANTY; without even the implied warranty of ## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU ## General Public License for more details. ## ## You should have received a copy of the GNU General Public License ## along with Octave; see the file COPYING. If not, see ## <http://www.gnu.org/licenses/>. ## -*- texinfo -*- ## @deftypefn {Function File} {} interpft (@var{x}, @var{n}) ## @deftypefnx {Function File} {} interpft (@var{x}, @var{n}, @var{dim}) ## ## Fourier interpolation. If @var{x} is a vector, then @var{x} is ## resampled with @var{n} points. The data in @var{x} is assumed to be ## equispaced. If @var{x} is an array, then operate along each column of ## the array separately. If @var{dim} is specified, then interpolate ## along the dimension @var{dim}. ## ## @code{interpft} assumes that the interpolated function is periodic, ## and so assumptions are made about the end points of the interpolation. ## ## @seealso{interp1} ## @end deftypefn ## Author: Paul Kienzle ## 2001-02-11 ## * initial version ## 2002-03-17 aadler ## * added code to work on matrices as well ## 2006-05-25 dbateman ## * Make it matlab compatiable, cutting out the 2-D interpolation function z = interpft (x, n, dim) if (nargin < 2 || nargin > 3) print_usage (); endif if (nargin == 2) if (isvector (x) && size (x, 1) == 1) dim = 2; else dim = 1; endif endif if (! isscalar (n)) error ("interpft: n must be an integer scalar"); endif nd = ndims (x); if (dim < 1 || dim > nd) error ("interpft: integrating over invalid dimension"); endif perm = [dim:nd, 1:(dim-1)]; x = permute (x, perm); m = size (x, 1); inc = 1; while (inc*n < m) inc++; endwhile y = fft (x) / m; k = floor (m / 2); sz = size (x); sz(1) = n * inc - m; idx = cell (nd, 1); for i = 2:nd idx{i} = 1:sz(i); endfor idx{1} = 1:k; z = cat (1, y(idx{:}), zeros (sz)); idx{1} = k+1:m; z = cat (1, z, y(idx{:})); z = n * ifft (z); if (inc != 1) sz(1) = n; z = inc * reshape (z(1:inc:end), sz); endif z = ipermute (z, perm); endfunction %!demo %! t = 0 : 0.3 : pi; dt = t(2)-t(1); %! n = length (t); k = 100; %! ti = t(1) + [0 : k-1]*dt*n/k; %! y = sin (4*t + 0.3) .* cos (3*t - 0.1); %! yp = sin (4*ti + 0.3) .* cos (3*ti - 0.1); %! plot (ti, yp, 'g', ti, interp1(t, y, ti, 'spline'), 'b', ... %! ti, interpft (y, k), 'c', t, y, 'r+'); %! legend ('sin(4t+0.3)cos(3t-0.1','spline','interpft','data'); %!shared n,y %! x = [0:10]'; y = sin(x); n = length (x); %!assert (interpft(y, n), y, 20*eps); %!assert (interpft(y', n), y', 20*eps); %!assert (interpft([y,y],n), [y,y], 20*eps); %!error (interpft(y,n,0)) %!error (interpft(y,[n,n]))