how to find duration of peak; starting and ending points

I've used findpeaks to isolate the largest peak in my signal. I wanted some help in finding the duration of the peak aka the starting and ending points of this peak, so that I could calculate area under curve? Right now Im having to visually inspect and pick arbitrary points to define this start and end points.

Answer 1

[Star Strider]

Experiment with ths ischange function, and if you have the Signal Processing Toolbox, the findchangepts function.

[Raheema Al Karim Damani]

Thank you for your response. I used findchangepts but since I have so much noise in my signal, I could not get the correct index. I've isolated this peak using the findpeak function:

[pks,loc,wid,p] = findpeaks(inv, 'SortStr','descend','NPeaks',1)

Can you advise on how I can apply findchangepts to only this peak. 

[Raheema Al Karim Damani]

here is the limits when I use 2 maxnumchanges

[Star Strider]

You will need to experiment with the functions. They cannot guess what you are thinking, and so they need to have some guidance.

[Raheema Al Karim Damani]

I appreciate your advice, but is there anyway to automate this process, instead of relying on findchangepts to create index for the peak because I have about 50 more plots.

[Star Strider]

In the plot you posted, there are several things findchangepts can consider to be change points.

The findchangepts function has a number of name-value pair arguments that you can use to tell it what you want it to do. The 'Statistic' and 'MinDistance' arguments are the ones I usually use. A different subset may be more appropriate for your purposes.

[Raheema Al Karim Damani]

Thank you your advice was helpful. The statistics arguments is really helpful in the findchangepts functions.

[Star Strider]

As always, my pleasure!

Answer 2

[Sindar]

widths are an optional output from findpeaks:

[pks,locs,widths,~] = findpeaks(data);

[Sindar]

Alternatively, you could potentially define the starting and ending points using findpeaks on the negative of your data.

[Raheema Al Karim Damani]

I believe that it finds the widths from the half prominence or half height, which would not accurately characterize the starting and ending points.

Also, I would like to find the index (x values) of the peak based on start and end points

Answer 3

[Alexis]

I had a similar issue, although I was working with biological signals. You could try the function I wrote to solve my problem: https://uk.mathworks.com/matlabcentral/fileexchange/81066-breathtimes?s_tid=srchtitle