Find specific points of increase within multiple different data sets

I am trying to write a function to identify the points as shown on this graph (not ones crossed with red).

However, I can't seem to get it to only identify the start and finish of the increase of data. Most of them are identifying points in between which are not wanted (but not at every point). This is what I have written so far:

function [Lpoints] = findpoints(Ldata)

%Calculate the point of abrupt changes in Ldata
Lchange = ischange(Ldata,'linear','Threshold',500);
%Find indices of changes 
Lindices=find(Lchange);
%Calculate change in L between these indices
figure;
plot(Ldata); 
xline(Lindices);

Lpoints=Lindices;

I have tried to exclude the central points by adding this (and adjusting the Ldiff>x parameter) to only get the start of the increase but this seems to eliminate the more subtle increases, and it would be preferred to get both points.

Lvalues = Ldata(Lindices);
Ldiff = [0; diff(Lvalues)];
number=Ldiff>40;
Ldiff(number)=NaN;
indices=~isnan(Ldiff);
Lindices2=Lindices(indices);
Lpoints=Lindices2

I have attached a selection of sample data, and it would be ideal to get something that works across the board. I may be taking the wrong approach, so any suggestions of how to go about this would be great!

Answer 1

[Mathieu NOE]

hello

try this... hope it helps - though I did not draw vertical lines but you have to start and ending points coordinates.

I tested the code on the 10 examples data provided

load('example_data.mat')
% whos
%   Name              Size            Bytes  Class     Attributes
% 
%   example1       1096x1              8768  double              
%   example10       486x1              3888  double              
%   example2       1327x1             10616  double              
%   example3       1116x1              8928  double              
%   example4        690x1              5520  double              
%   example5        642x1              5136  double              
%   example6        720x1              5760  double              
%   example7        834x1              6672  double              
%   example8        655x1              5240  double              
%   example9        790x1              6320  double 

data = example1;

% high pass filter the data first
[b,a] = butter(2,0.01,'high');
datad = filtfilt(b,a,data);

% find start ,  end points = positive and negative peaks of the high pass
% filtered data (with islocalmax)

minpeakdist = 50;
min_amplitude= max(abs(datad))/10;

tf=islocalmax(datad,'MinSeparation',minpeakdist, 'MinProminence', 1);
locsp = find(tf);
locsp = locsp(datad(locsp)>min_amplitude);
tf=islocalmax(-datad,'MinSeparation',minpeakdist, 'MinProminence', 1);
locsn = find(tf);
locsn = locsn(datad(locsn)<-min_amplitude);

locsn(locsn>locsp(end)) = []; % remove "start" point wich is now followed by "end" point at the end of the data

% plot
xaxis = 1:numel(data);
% plot(xaxis,datad,xaxis(locsn),datad(locsn),'dk',xaxis(locsp),datad(locsp),'dr');
plot(xaxis,data,xaxis(locsn),data(locsn),'dk',xaxis(locsp),data(locsp),'dr');
legend('data','start points','end points','Location','NorthEastOutside');

[pkll201]

That's amazing - works perfectly! Thank you so much :)

[Mathieu NOE]

my pleasure !

have a great day