Is something like this what you're trying to do?
load VEX.mat
% plot VEX
figure
plot(VEX,'b.')
First, identity where VEX has large jumps up or down, from one segment to the next:
jump_threshold = 10;
idx = [true; abs(diff(VEX))>jump_threshold; true];
idx = find(idx)
and plot (for reference) the starts (in green) and ends (in red) of the separate segments:
figure
plot(VEX,'b.')
hold on
plot(idx(1:end-1),VEX(idx(1:end-1)),'g.')
plot(idx(2:end)-1,VEX(idx(2:end)-1),'r.')
Now identify the value (and corresponding index of VEX) in each segment nearest to each of 8, 12, 18, 22, 58, 62, 68, and 72:
vals = [8 12 18 22 58 62 68 72];
NV = numel(vals);
NI = numel(idx)-1;
all_idx = NaN(NI,NV);
all_vals = NaN(NI,NV);
for ii = 1:NI
% subVEX is the data for the ii-th segment:
subVEX = VEX(idx(ii):idx(ii+1)-1);
for jj = 1:NV
% find the element nearest vals(jj) in segment ii
[d,min_idx] = min(abs(subVEX-vals(jj)));
% only store it if it is closer than jump_threshold away from vals(jj)
% (e.g., don't want to match the bottom point of a segment that is
% entirely above 50 to the value 22):
if d <= jump_threshold
all_idx(ii,jj) = min_idx+idx(ii)-1;
all_vals(ii,jj) = subVEX(min_idx);
end
end
end
This is what the indices and VEX values at those points are. (NaNs occur where the nearest point in a segment was further than jump_threshold away from vals(jj)):
disp(all_idx)
disp(all_vals)
Plot the non-NaN elements of those:
figure
plot(VEX,'b.')
hold on
to_plot = ~isnan(all_idx);
plot(all_idx(to_plot),all_vals(to_plot),'r*')
Note that your approach of using find(__,1,'first') and find(__,1,'last') would never work on the interior segments since you are searching in the entire VEX vector. That's why I break VEX into segments first and search each segment individually.
