Trying to figure out a algorithm to filter out some data, since I dont have any additional toolbox.

Hi,

I have uploaded a mat file which contains two vectors e and v. e is for energy and v is for voltage, but thats not important. And I plot them like this.

plot(v,e);  
grid on;
grid minor;
xlim([3.485,3.6]);
set(gca, 'XDir','reverse');

and when you plot the values, you get this plot with those spikes. I am trying to figure out how to remove those spikes, but i cant see any patterns. I dont have any toolbox, so i have to use matlab and my brain to figure this out. But i am currently struggling to come up with an idea. If anyone can help me, kindly please help me. Thank you.

Answer 1

[Star Strider]

The filloutliers function is a core MATLAB function (no toolboxes required) introduced in R2017a.

Using it on your data --

LD = load('eeandv.mat');
e = LD.e;
v = LD.v;

figure
plot(v,e);  
grid on;
grid minor;
xlim([3.485,3.6]);
set(gca, 'XDir','reverse');
xlabel('v')
ylabel('e')
title('Original')

●

e = filloutliers(e, 'linear','movmedian',10);

figure
plot(v,e);  
grid on;
grid minor;
xlim([3.485,3.6]);
set(gca, 'XDir','reverse');
xlabel('v')
ylabel('e')
title('Using ''filloutliers''')

.

[Govind Sankar Madhavan Pillai Ramachandran Nair]

thank you

[Star Strider]

As always, my pleasure!

Answer 2

[Torsten]

load("eeandv.mat")
tol = 0.005;
idx = abs(diff(e)) <= tol;
e = e(idx);
v = v(idx);
plot(v,e)
grid on;
grid minor;
xlim([3.485,3.6]);
set(gca, 'XDir','reverse');

●

[Govind Sankar Madhavan Pillai Ramachandran Nair]

thank you

Answer 3

[Alan Stevens]

Here's one possibility, though it isn't a general solution, it just works for this data set

load('eeandv.mat')
xlo =3.485; xhi = 3.6; 
ixlo = find(v==xlo); ixhi= find(v==xhi);
ylo = e(ixlo); yhi = e(ixhi);
x = [xlo, xhi];
y = @(x) ylo + (x-xlo)*(yhi-ylo)/(xhi-xlo);

e(e<y(v)) = NaN;
plot(v,e,'.'); 
grid on;
grid minor;
xlim([3.485,3.6]);
set(gca, 'XDir','reverse');

[Govind Sankar Madhavan Pillai Ramachandran Nair]

thank you