finding the duration of signal peak period

HI below image represents data from video frames and has 4 peaks, the (x-axis) is the frame numbers, (y-axis) is the data values . how the duration of each peak can be found in seconds ?

Answer 1

[Star Strider]

Use the findpeaks function on the negative of your signal, and experiment with the name-value pair options, such as 'WidthReference'. Then have it return all 4 outputs:

[pks,locs,w,p] = findpeaks(-data, ... );

[sarmad m]

these features are from video has 30 FPS

is WidthReference returns the duration of peak signal width ?

[Star Strider]

I do not understand what you are asking.

See the documentation on the findpeaks function to understand how to use it and what it returns.

[Image Analyst]

We don't care what form your original data took. All that matters is that you somehow processed it to get a 1-D signal, what Star called data. As a matter of fact, it looks like, from your plot with valleys labeled, that you've already called findpeaks(). Find find peaks returns indexes of the peaks (or valleys), all you have to do is to find the distance between indexes and then multiply by the time per frame:

peak1Width = (index2-index1) * frameTime;

[sarmad m]

Dear Star Strider ,I meant is WidthReference returns the duration of below peak for example ?

Image Analyst:

is peak1Width = (index2-index1) * frameTime; returns the duration between the peaks or single peak duration ?

[Star Strider]

The findpeaks function would have as difficult a time as I would have defining ‘duration’. I would go with one of its 'WidthReference' options and leave it at that.

Otherwise, since your ‘peaks’ seem to be lognormal (or log-Gaussian) rather than normal (or Gaussian), deciding when they begin and end would be difficult. It would probably be best to model the peak contour as a log-Gaussian function, then identify and fit each ‘peak’, and use those parameters to characterise your data. (This approach seems unnecessarily complicated to me, unless you have a specific reason to need those estimated parameters. However, estimating them would be necessary if this were a chromatography image or some such and you wanted to estimate the total concentration under each peak.)

[sarmad m]

I need a simple solution , I will use WidthReference .

In the help says that :

[___] = findpeaks(data,Fs) specifies the sample rate, Fs, of the data.

where FS is Sample rate, specified as a positive scalar. The sample rate is the number of samples per unit time.

if my video is 30fps and I use this

AV=-AV;
[pks_smoothed,locs_smoothed,widths_smoothed,proms_smoothed] = findpeaks(AV,30,'MinPeakProminence',0.05);

is widths_smoothed will return the duration ?

Regards

[Star Strider]

The widths will return in the dimensions you provide to findpeaks, or in terms of the indices if you do not provide it with a time vector or sampling frequency scalar. (One function — and I have now forgotten which one — will not return the correct output with only the sampling frequency but will if you give it the full time vector.) If giving it the sampling frequency does not work, give it the entire time vector instead (in the same position in the argument list). That always works for me.

[sarmad m]

In my example , the video type is .avi, 30fps and 4 seconds the output features are here : Data

is sampling rate =4/(1/30)=121.21 ?

is this correct ?

[pks_smoothed,locs_smoothed,widths_smoothed,proms_smoothed] = findpeaks(AV,121.21,'MinPeakProminence',0.05);

how time vector can be used in this case ?

[Star Strider]

I am not certain what your data actually represent. The ‘units’ of your peak locations and widths will be in terms of whatever you specify. If you specify nothing, they will be in terms of indices of your data vector. If your x-axis is time, ‘locs_smoothed’ and ‘widths_smoothed’ would be in units of time. The ‘proms_smoothed’ units will be in whatever your y-axis is.

If each plotted data point in your plot is some point (or a point representing some data in the full frame) in what I assume is your video, and those are presented at 30 fps, then your sampling frequency, ‘Fs’, is 30 Hz. The number of frames you have is not important with respect to your sampling frequency.

If the x-axis in your plot are indices, you can substitute your time vector directly to get it and all values derived from it in units of time.

Answer 2

[Brian Mabe]

I am having a similar problem with the following: I need to calculation the duration and number of peaks in the data plotted by the given code:

clear clc load Time_X.txt load Turbid_Y.txt x=Time_X; y=Turbid_Y; t=[x,y]; i=1; a=0; b=0; while b==0 if t(i,2)>=0 if y(i)>=55 a=a+1; % disp(y(i)) value(a)=y(i); end i=i+1; end if i==14513 b=1; end

end

scatter(x,y,'.') hold on plot([0 31], [55 55]) xlim([0 31]); ylim([0 190]); if y>55 disp('y') end