Matlab - Funcion findpeaks de matlab

function [pks,locs] = findpeaks(X,varargin)

%FINDPEAKS Find local peaks in data

%   PKS = FINDPEAKS(X) finds local peaks in the data vector X.

%

%   [PKS,LOCS]= FINDPEAKS(X) also returns the indices LOCS at which the

%   peaks occur.

%

%   [...] = FINDPEAKS(X,'MINPEAKHEIGHT',MPH) finds only those peaks that

%   are greater than MINPEAKHEIGHT MPH. Specifying a minimum peak height

%   may help in reducing the processing time. MPH is a real valued scalar.

%   The default value of MPH is -Inf.

%

%   [...] = FINDPEAKS(X,'MINPEAKDISTANCE',MPD) finds peaks that are at

%   least separated by MINPEAKDISTANCE MPD. MPD is a positive integer

%   valued scalar. This parameter may be specified to ignore smaller peaks

%   that may occur in close proximity to a large local peak. For example,

%   if a large local peak occurs at index N, then all smaller peaks in the

%   range (N-MPD, N+MPD) are ignored. If not specified, MPD is assigned a

%   value of one. 

%

%   [...] = FINDPEAKS(X,'THRESHOLD',TH)finds peaks that are at least

%   greater than their neighbhors by the THRESHOLD TH. TH is real valued

%   scalar greater than or equal to zero. The default value of TH is zero.

%

%   [...] = FINDPEAKS(X,'NPEAKS',NP) specifies the number of peaks to be

%   found. NP is an integer greater than zero. When specified, the

%   procedure terminates once NP peaks are found and the NP peaks are

%   returned. If not specified, all peaks are returned.

%

%   [...] = FINDPEAKS(X,'SORTSTR',STR) specifies the direction of sorting

%   of peaks. STR can take values of 'ascend','descend' or 'none'. If not

%   specified, STR takes the value of 'none' and the peaks are returned in

%   the order of their occurrence.

%

%   See also DSPDATA/FINDPEAKS

%   Copyright 2007-2009 The MathWorks, Inc.

%   $Revision: 1.1.6.10 $  $Date: 2009/10/16 06:41:09 $

error(nargchk(1,11,nargin,'struct'));

validateattributes(X,{'numeric'},{'nonempty','real','vector'},...

    'findpeaks','X');

% Check the input data type. Single precision is not supported.

try

    chkinputdatatype(X);

catch ME

    throwAsCaller(ME);

end

%#function dspopts.findpeaks

hopts = uddpvparse('dspopts.findpeaks',varargin{:});

Ph  = hopts.MinPeakHeight;

Pd  = hopts.MinPeakDistance;

Th  = hopts.Threshold;

Np  = hopts.NPeaks;

Str = hopts.SortStr;

pks = [];

locs = [];

M = numel(X);

if (M < 3)

    datamsgid = generatemsgid('emptyDataSet');

    error(datamsgid,'Data set must contain at least 3 samples.');

else

    Indx = find(X > Ph);

    if(isempty(Indx))

        mphmsgid = generatemsgid('largeMinPeakHeight');

        warning(mphmsgid,'Invalid MinPeakHeight. There are no data points greater than MinPeakHeight.');

    else

        % validate value of Pd and set default values for Pd and Np

        [Pd,Np] = setvalues(Pd,Np,M);

        if(Pd >= M)

            pdmsgid = generatemsgid('largeMinPeakDistance');

            error(pdmsgid,'Invalid MinPeakDistance. Set MinPeakDistance as an integer in the range between 1 and %s.',...

                num2str(M));

        else

            [pks,locs] =getpeaks(X,Indx,Pd,Th,Np);

        end

    end

end

if isempty(pks)

    npmsgid = generatemsgid('noPeaks');

    warning(npmsgid,'No peaks found.')

elseif(~strcmp(Str,'none'))

    [pks,s]  = sort(pks,Str);

    if(~isempty(locs))

        locs = locs(s);

    end

end

%--------------------------------------------------------------------------

function [Pd,Np] = setvalues(Pd,Np,L)

if ~isempty(Pd) && (~isnumeric(Pd) || ~isscalar(Pd) ||any(rem(Pd,1)) || (Pd < 1))

    Nmsgid = generatemsgid('invalidMinPeakDistance');

    error(Nmsgid,'MinPeakDistance should be an integer greater than 0.');

end

if(isempty(Pd))

    Pd = 1;

end

if(isempty(Np))

    Np = L;

end

%--------------------------------------------------------------------------

function [pks,locs] =getpeaks(Data,Indx,Pd,Th,Np)

% This function finds peaks in data set Data whose index set Indx is

% disjoint. Some data points were removed from the original set through

% preprocessing

m = 0;                  % counter for peaks found

L = numel(Indx);

M = numel(Data);

% Pre-allocate for speed

pks  = zeros(1,Np);

locs = zeros(1,Np);

endindx = M;      % last point in the index set

j = 0;

% First, the "Pd" neighbors, on either side, of the current data point are

% found. Then the current data point is compared with these neighbors to

% determine whether it is a peak.

while (j < L) && (m < Np)

    j = j+1;

    currentIdx = Indx(j);

    % Update the leftmost neighbor index if there is a peak within "Pd"

    % neighbors of leftmost index

    if(m > 0)

        prevPeakBoundR = min([locs(m)+Pd, endindx-1]);

        if currentIdx < prevPeakBoundR

            k = find(Indx(j+1:end) > prevPeakBoundR,1,'first');

            if ~isempty(k)

                j = j+k;

                currentIdx = Indx(j);

            else

                break;

            end

        end

    end

    % leftmost neighbor index

    endL = max(1,currentIdx - Pd);

    % rightmost neighbor index

    endR = min(currentIdx + Pd,endindx);

    % create neighbor data set

    temp = Data(endL:endR);

    % set current data point to -Inf in the neighbor data set

    temp(currentIdx-endL+1) = -Inf;

    % Compare current data point with all neighbors

    if(all((Data(currentIdx) > temp+Th)) && (currentIdx ~=1)&& (currentIdx~=M))

        m = m+1;

        locs(m) = currentIdx;  % loctions of peaks

        pks(m)  = Data(currentIdx);  % peaks

    end

end

% return all peaks found

if m~=0

    locs = locs(locs > 0);

    pks  = pks(1:length(locs));

else

    locs = [];

    pks = [];

end

% [EOF]