Unexpected Result When Using Sinc Interpolation

Question

blue is how I tried to sinc interpolate. why would something like this happen?

Maybe do it in Frequency Domain - https://dsp.stackexchange.com/questions/25131. — Mark, Sep 01 '21 at 08:06

score 3 · Accepted Answer · edited Sep 01 '21 at 08:16

3

Since Sinc based Interpolation requires you to know the data at any point. Hence it is not feasible.

You might do a Truncated Sinc Interpolation.
The artifacts you're seeing can be caused by a kernel which is too short or the parameters aren't good.

In order to create a good Sinc kernel you need to know things about the Band Width of the signal and the Sampling Rate.
Did you took those into account?

edited Sep 01 '21 at 08:16

Mark

357
1
6
24

answered Aug 09 '15 at 11:01

Royi

19,608
4
197
238

I just zoomed to a part of the signal. the signal goes to zero and it is sampled often enough – Gappy Hilmore Aug 09 '15 at 11:06
Could you post the MAT file of the data and the M file you are running? – Royi Aug 09 '15 at 11:18
http://pastebin.com/sLYWrH5q – Gappy Hilmore Aug 09 '15 at 13:32
I tried extending the sinc I convolved my samples with 100 times, it gave acceptable results. – Gappy Hilmore Aug 09 '15 at 13:37
Dra, i don't think the problem is from not choosing good parameters. grd just ain't doing it right. looks like the OP might be using $|\operatorname{sinc}(\cdot)|$ instead. – robert bristow-johnson Aug 10 '15 at 00:49
I couldn't access his code, is it there? – Royi Aug 10 '15 at 05:35
@Drazick http://pastebin.com/aEwDv012 this should be available – Gappy Hilmore Aug 10 '15 at 11:46
@Drazick i posted it as an answer – Gappy Hilmore Aug 10 '15 at 13:57
1

How boundaries are handled is also important aspect. – Eric Johnson Feb 13 '22 at 21:03

score 0 · Answer 2 · answered Aug 10 '15 at 13:56

t=linspace(-.5,.5,256);
x=exp(-pi*t.^2*16).*(sin(2*pi*40*t)+0.154*cos(2*pi*47*t)-1.454*cos(2*pi*27*t));
figure;plot(t,x)

tt=linspace(-.5,.5,256*8-7);
xorj=exp(-pi*tt.^2*16).*(sin(2*pi*40*tt)+0.154*cos(2*pi*47*tt)-1.454*cos(2*pi*27*tt));
figure;plot(tt,xorj)

xf=SincInt(x,8,1);
sh=0;
xf=[ xf(1+sh:end) zeros(1,sh)];
figure;plot(tt,xf)
hold on;plot(tt,xorj)
figure;plot(tf,xf-xorj)

function f = SincInt( f,k ,varargin)
% function f = SincInt( f,k ,varargin)
% varargin=1 to keep the beginning and the end the same

nargin=length(varargin);

N=length(f);
f=[zeros(1,N*(k-1)) f];
for  i = 1: N
    f(k*i-k+1:k*i)= [zeros(1,k-1) f(N*(k-1)+i)];
end

f=conv(f,sinc(-60:1/k:60),'same');

if(nargin==1 && varargin{1})
    f=[f(k:length(f)) ];
end

end

Unexpected Result When Using Sinc Interpolation

2 Answers2