Finding points of signal

I have some data labelled x_v[n] and I am trying to locate the starting and ending points of the peaks from my data.

Till now I obtained x_v[n] and then I performed template matching.

I require to find the starting and ending points of the peaks on x_v[n] Any idea as to how I can do that please?

Here is my code till now:

fs = 256;                          
N = length(EOG_Data);             
T = 1/fs;                         
Time = [0:(N - 1)] * T;         

figure();
plot(Time, EOG_Data);
title('x_v[n]');

[pks, locs] = findpeaks(R, 'MinPeakHeight',10e6); %finding peaks
figure();
plot(Time, R); 
hold on;
plot(EOG_Time(locs), pks, '*r') 
hold off;

Answer 1

[Image Analyst]

Well you could start at a peak and go forward and backward until the signal begins to turn upward.

Or you could just use the widths returned by findpeaks(), which is about half way up. I have a demo for this:

clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clear;  % Erase all existing variables. Or clearvars if you want.
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 22;
fprintf('Beginning to run %s.m ...\n', mfilename);

EOG_Data = load('EOG_Data.mat');
EOG_Data_v = double(EOG_Data.EOG_Data_4ch(1,:)) - double(EOG_Data.EOG_Data_4ch(2,:));

fs = 256;
N = length(EOG_Data_v);
T = 1/fs;
EOG_Time = [0:(N - 1)] * T;

subplot(2, 1, 1);
plot(EOG_Time, EOG_Data_v);
title('x_v[n]', 'FontSize', fontSize);
grid on; 

Blink = EOG_Data_v(500:720); %template

[R, lags] = xcorr(EOG_Data_v, Blink); %cross-correlation
R = R(lags>=0);
lags = lags(lags>=0);

[peakValues, indexesOfPeaks, widths, proms] = findpeaks(R, 'MinPeakHeight',10e6); %finding peaks
subplot(2, 1, 2);
plot(EOG_Time, R);
title('Occurance of Blinks from Cross-Correlation plot', 'FontSize', fontSize);
xlabel('Time in seconds', 'FontSize', fontSize); 
ylabel('Voltage in mV', 'FontSize', fontSize); 
grid on; 
axis tight;
hold on;
plot(EOG_Time(indexesOfPeaks), peakValues, '*r')
% Plot widths
valleyIndexes = zeros(length(widths), 2);
for k = 1 : length(widths)
	% Get the indexes.
	valleyIndexes(k, 1) = floor(indexesOfPeaks(k) - widths(k)/2);
	valleyIndexes(k, 2) = ceil(indexesOfPeaks(k) + widths(k)/2);
	% Get the times.
	t1 = EOG_Time(valleyIndexes(k, 1));
	t2 = EOG_Time(valleyIndexes(k, 2));
	% Draw vertical lines on the plot.
	xline(t1, 'Color', 'm', 'LineWidth', 2);
	xline(t2, 'Color', 'm', 'LineWidth', 2);
end
hold off;
% Maximize figure window.
g = gcf;
g.WindowState = 'maximized'
fprintf('Done running %s.m.\n', mfilename);

[Image Analyst]

By the way, what does EOG stand for?

[Rudy Steel]

Electrooculogram.

Thank you for the reply, however I require to find the starting and ending points on the x_v[n] plot, I applied your code to that plot but I am not obtaining the time slots required with t1 and t2. Basically I require something similar to this:

As afterwards I need to drop the voltage levels between those two points to zero and conjoin the gap so that I would end up with a plot without those peaks.

[Star Strider]

Electrooculogram

[Image Analyst]

Then I'd use my method of starting at each peak and "falling down" each side until the signal starts increasing again. Hopefully you know how to do that. You'll have 2 for loops (one for each side) inside a for loop over all the peaks. Let me know if you can't figure out how to do it.

I tried first to do it by inverting the signal and finding peaks on either side but I wasn't getting anywhere with that approach.

[Rudy Steel]

So, I tried to do as you suggested using a for loop that iterates over all the peaks then I used two if statements to find the starting and ending point however I am not able to arrive to what I need. Any help would be appreciated

startingpoint = zeros(size(pks)); % pks are the peakValues
endingpoint = zeros(size(pks));

for i = 1:size(pks,1)
    sp = find((EOG_Data_v < pks(i)));
    if isempty(sp)
        sp = 1;
    end
    startingpoint(i) = sp;
    ep = find((EOG_Data_v < pks(i)));
    if isempty(ep)
        ep = 1;
    end
    endingpoint(i) = ep;
end

[Image Analyst]

Oh, sorry to hear that. When I did it, it worked pretty well, I thought, though I'm not exactly sure where you think the "base" of the peaks in R should be. But anyway, try this:

clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
clear;  % Erase all existing variables. Or clearvars if you want.
workspace;  % Make sure the workspace panel is showing.
format long g;
format compact;
fontSize = 22;
fprintf('Beginning to run %s.m ...\n', mfilename);

EOG_Data = load('EOG_Data.mat');
EOG_Data_v = double(EOG_Data.EOG_Data_4ch(1,:)) - double(EOG_Data.EOG_Data_4ch(2,:));

fs = 256;
N = length(EOG_Data_v);
T = 1/fs;
EOG_Time = [0:(N - 1)] * T;

subplot(2, 1, 1);
plot(EOG_Time, EOG_Data_v);
title('x_v[n]', 'FontSize', fontSize);
grid on; 

Blink = EOG_Data_v(500:720); %template

[R, lags] = xcorr(EOG_Data_v, Blink); %cross-correlation
R = R(lags>=0);
lags = lags(lags>=0);

[peakValues, indexesOfPeaks, widths, proms] = findpeaks(R, 'MinPeakHeight',10e6); %finding peaks
subplot(2, 1, 2);
plot(EOG_Time, R);
title('Occurance of Blinks from Cross-Correlation plot', 'FontSize', fontSize);
xlabel('Time in seconds', 'FontSize', fontSize); 
ylabel('Voltage in mV', 'FontSize', fontSize); 
grid on; 
axis tight;
hold on;
plot(EOG_Time(indexesOfPeaks), peakValues, '*r')
% Plot widths
valleyIndexes = zeros(length(widths), 2);
for k = 1 : length(widths) % For each peak...
	% Get the indexes.
	thisPeakIndex = indexesOfPeaks(k);
	% Fall down the left side of this peak.
	for k2 = (thisPeakIndex - 1) : -1 : 2
		if R(k2) < R(k2-1)
			leftPeak(k) = k2-1;
			break;
		end
	end
	% Fall down the right side of this peak.
	for k2 = (thisPeakIndex + 1) : length(R)
		if R(k2) > R(k2-1)
			rightPeak(k) = k2-1;
			break;
		end
	end
	% Get the times.
	t1 = EOG_Time(leftPeak(k));
	t2 = EOG_Time(rightPeak(k));
	% Draw vertical lines on the plot.
	xline(t1, 'Color', 'm', 'LineWidth', 2);
	xline(t2, 'Color', 'm', 'LineWidth', 2);
end
hold off;
% Maximize figure window.
g = gcf;
g.WindowState = 'maximized'
fprintf('Done running %s.m.\n', mfilename);

If you'd rather do it on EOG_Data_v instead of R, you could of course do that with the obvious modifications, though you might need to smooth EOG_Data_v a bit or put some threshold on how much the signal has to turn upwards.

[Rudy Steel]

Your first response was actually good I had a mix up in the variables but now it is working as intended. I changed xline to actual plots having '*r*' and I also managed to erase the data between the starting and ending points and get my desired result. Thanks a lot!