MATLAB: How to locate peak frequencies using “find peak” and add it to an existing FFT

fftfind peak

My friends and I are experimenting at our private lab and using a MatLab Fast Fourier Transform program. We would like to add a "find peak" function to locate one or more locations on the FFT. How would I insert this function? Thanks.

Best Answer

There are some File Exchange contributions that do this.

The Signal Processing Toolbox has the findpeaks function, and since you’re doing EEG analysis, you’ll be using other signal processing techniques such as filtering and spectrographic analysis. so it’s worth getting it if you don’t already have it.

Related Solutions

MATLAB: Help locating peaks

"values of the actual bins" is not well-defined here. Do you mean the frequency that corresponds to the bin number?

If so then as Wayne King posted earlier this year,

the DFT bins are spaced at intervals of Fs/N where N is the length of the DFT vector and Fs is the sampling frequency. The first bin is zero frequency.

Thus, ((sampling frequency) times (fft bin number minus one) divided by (length of fft result)) gives frequency associated with the bin number.

MATLAB: Locating Peaks Naively

In case you don't have the Signal Processing Toolbox, or its findpeaks function is not general enough, you could try the following function - though I don't claim it's the the fastest way possible!

function peaks = peakfind(v, x)
%PEAKFIND finds local peaks
%   PEAKS = PEAKFIND(V, X) returns the indices of elements of vector V such
%   that V(P) is greater than V(P + K) for all integers K such that K >=
%   -X, K <= X and K ~= 0. Only peaks for which P > X and P < length(V)+1-X
%   are returned.
%
%   V must be a row or a column vector. X must be a positive integer. The
%   result is a vector containing the indices P.
% Argument checking omitted for speed.
pks = true;
va = v(x+1 : end);
for k = 1:x
   vb = v(x+1-k : end-k);
   pksa = va > vb;
   pksb = vb > va;
   pks = pks & pksa(1 : end-x) & pksb(1+k : end-x+k);
end
peaks = find(pks) + x;
end

One thing to note: if you define a peak to be an element that is greater than all its neighbours, repeated values in the input vector will never be seen as a peak. So [ 1 100 100 1] (with X=1) has no peaks using your definition. If you want to find such plateaux, you can do it by breaking the symmetry and finding one of the values. In the code above, you could replace "pksb = vb > va" with "pksb = ~pksa" (and amending the description) to achieve this.

Best Answer

Related Solutions

MATLAB: Help locating peaks

MATLAB: Locating Peaks Naively

Related Question