Digitization of noisy signal

I have a signal as shown in the image here. I am trying to digitize the signal so that it catches the right jump only. My system is jumping between two states with noise. I want to calculate how much time it stays up and how much down. Could you please bring a super cool idea? Thanks for your help

Answer 1

[Star Strider]

If your have version R2016a or later of the Signal Processing Toolbox, use the findchangepts (link) function.

[suman Dhamala]

Hi Sar Strider, I have it and realize it is unable to solve it. I am looking for a good way to write my own code. I need help on this sooner

[Star Strider]

Another option is the cusum (link) function.

[suman Dhamala]

Thanks, Sar strider, To use findchangepts(), at first I need to look the nature of my data cycle and give the argument. It looks it takes a lot of time to analyse it.

[Star Strider]

As always, my pleasure.

Without your data, I cannot write specific code for your problem. The findchangepts() function has several name-value pair arguments that can simplify your analysis.

[suman Dhamala]

This code generates the data instantly.

phi(1)=0.66702; phi(2)= 0.667003; phidc(1)=0;phidc(2)= 0.00001;

for n=2:10^5 ;

p= 0.5.*rand-0.25; q = 1.*rand -0.5 ;

phi(n+1)= 0.0031+ 1.9676*phi(n) -0.9686*phi(n-1)- 0.0039*cos( 0.5*phidc(n))* sin(phi(n)) + 0.0390* q ;

phidc(n+1)= 1.8873*phidc(n) -0.9686*phidc(n-1)- 0.0079*sin(0.5*phidc(n))* cos(phi(n)) + 4*0.0390* p ;

end

figure(1110)

plot(phi)

[Star Strider]

Good to know.

However, it does not produce the data you posted.

[suman Dhamala]

Hi Star Strider,

I also tried it and figure out the problem was on the decimal number. I fixed it, can you please try it. Thanks

phi(1)=0.66702; phi(2)= 0.667003; phidc(1)=0;phidc(2)= 0.00001;

for n=2:10^5 ;

p= 0.5.*rand-0.25; q = 1.*rand -0.5 ;

phi(n+1)= 0.0031+ 1.9676*phi(n) -0.9686*phi(n-1)- 0.0039*cos( 0.5*phidc(n))* sin(phi(n)) + 0.0390* q ;

phidc(n+1)= 1.8873*phidc(n) -0.9686*phidc(n-1)- 0.0079*sin(0.5*phidc(n))* cos(phi(n)) + 4*0.0390* p ;

end

figure(1110)

plot(phi)

[Star Strider]

Try this:

t = 1:numel(phi);
ipt = findchangepts(phi, 'MinThreshold',1000, 'Statistic','std');

figure(1110)
plot(t, phi)
hold on
plot([t(ipt(1:2:end)); t(ipt(1:2:end))], [zeros(1,numel(ipt(1:2:end))); 5*ones(1,numel(ipt(2:2:end)))], '-r', 'LineWidth',1.5)
hold off

It plots vertical lines at the transitions.