find peaks

does anyone know, is there code for finding peaks in a matrix that have noise? (because of this noise i cant use findpeaks command in matlab) the noise has so small amplitude in compared with amplitude of main wave. thanks

Answer 1

[Fangjun Jiang]

If you have noise in the data, you need to filter it first. See:

doc filter

[mohammad]

i funded some code for analog with noise that finds peaks of these waves, but i need these cods for discrete form (matrix)

[Fangjun Jiang]

FILTER is One-dimensional digital filter.

FILTER2 is Two-dimensional digital filter

[mohammad]

thanks dear Fangjun, in doc filter there is a parameter b=numerator coefficient vector, how must i calculate this?

[Fangjun Jiang]

You don't. That is the parameter to specify your designed filter. See the example in the doc.

Answer 2

[James Tursa]

You might try one of the FEX submissions, e.g. PeakFinder by Nate Yoder:

http://www.mathworks.com/matlabcentral/fileexchange/25500-peakfinder

[mohammad]

dear James thanks for replying

my file is a .xls file. could i send this for you and help how can i use this code for this file?

[Fangjun Jiang]

use xlsread() to read the data into MATLAB workspace first, then apply filter.

Answer 3

[Honglei Chen]

Hi Mohammad,

Can you explain why you cannot use findpeaks? I think you can specify MinPeakHeight to avoid false alarm from the noise.

Thanks.

[mohammad]

hi dear Hongle,thanks for replying

when i use findpeaks, because of noise it gives me 1000 peaks but the matrix has only 5 0r 10 desire peaks, and when i want to sort these peaks and gather 5 or 10 biggest at the first of sorting, again because of noise it gives the biggest peaks and some values (for 4 or 9 others peaks) around that biggest peak. i designed some filters for this but aren't accurate. i saw some nice codes for finding peaks of analog signals with noise but my data is in matrix and discrete form. i think there are some accurate codes for matrix form, saving my time.

Answer 4

[Harry MacDowel]

Sometimes identifying the noise would be a good start. Try Allan Variance method.

Good luck!