Introduction
Hi!
This is my first rst post!
And this is a link to a page
But does it format Matlab code?
>> y = pitchfilter(d,sr);
>> soundsc(y,sr)
>> soundsc(d,sr)
>> soundsc(y,sr)
>> y = pitchfilter(d,sr);
>> y = pitchfilter(d,sr);
>> soundsc(y,sr)
>> max(abs(y))
ans =
0.1631
>> max(abs(d))
ans =
0.2468
>> wavwrite(y, sr, '~/Downloads/progressonhistogramequalization/nr1ex_w+pitchfilt.wav');
>> print -dpdf ~/Downloads/progressonhistogramequalization/nr1ex_w+pitchfilt.pdf
We will see!
Here's an actual matlab function;
function Y = map_vals(X,vmap)
% Y = map_vals(X,vmap)
% X is a vector, map is two rows indicating input, output pairs
% 2014-04-29 Dan Ellis dpwe@ee.columbia.edu
% Just run on a vector
% figure slopes
gap = diff(vmap, 1, 2);
% repeat final gap on both src and dst rows
gap = [gap, gap(:,end)];
% Don't allow zero (or negative?) gaps
gap(find(gap <= 0)) = eps;
% do mapping. 1+ for Matlab indexing later on
Xix = 1 + max(0, sum(bsxfun(@gt, X', vmap(1,:)), 2)-1);
Xdelta = (X - vmap(1, Xix))./gap(1, Xix);
Y = vmap(2, Xix) + Xdelta.*gap(2, Xix);
How about that?
Does it do nicer with python?
def map_vals(X, vmap):
""" X is a vector, map is two rows indicating input, output pairs """
if len(np.shape(X)) > 1:
# If X is a matrix, run separately on each row
return np.array([map_vals(XX, vmap) for XX in X])
else:
# Just run on a vector
# figure slopes
gap = np.diff(vmap)
# repeat final gap on both src and dst rows
gap = np.c_[gap, gap[:,-1][:, np.newaxis]]
# Don't allow zero (or negative?) gaps
gap[np.nonzero(gap <= 0)] = np.finfo(float).eps
# do mapping
Xix = np.maximum(0,
np.sum(np.greater.outer(X, vmap[0]), axis=1)
- 1 )
Xdelta = (X - vmap[0, Xix])/gap[0, Xix]
return vmap[1, Xix] + Xdelta*gap[1, Xix]
So that's a thing.
