[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Finding peaks/max in a graph

From: edA-qa mort-ora-y
Subject: Finding peaks/max in a graph
Date: Mon, 05 Apr 2004 12:10:51 +0200
User-agent: Mozilla Thunderbird 0.5 (X11/20040208)

I need to find the peaks in a graph (sampled data). By peaks I mean each point at which the slope changes from positive to negative (or 0 slope, but these are usually cusps).

As this is sampled data (resulting from an FFT actually), I need to find only a certain number of the peaks (the maximum peaks, since the noise in the sample yields a high number of real peaks).

I wrote a function which does this, but it is extremely slow (I'm working with sample sets of 100K in size). Does anybody else know how I can achieve this any faster, or which standard functions.

Purpose: To find the most predominant frequencies in a waveform (sample fed into FFT, then into peaks finding algorithm)

Attached: The script I wrote

edA-qa mort-ora-y
Idea Architect

function p = find_peaks( data, num )

        cur = data(1);
        mn = cur;
        dir = 0;

        peaks = [];

        for i = 2:length(data)
                s = sign( data(i)-cur );
                if( s != dir )
                        dir = s;

                        if( dir < 0 )
                                peaks   = [peaks; i, data(i)];

                cur = data(i);
                mn = min(cur,mn);

        l = rows(peaks);
        if( num > l )
                num = l;

        # Ensure we are using a matrix, otherwise sort treats it as a vector
        # we should populate this with the minimum value actually... (TBD)
        if( l == 1 )
                peaks = [peaks; 0, mn];
                l = 2;

        [ spv, spi ] = sort( peaks );

        p = [];
        for i = 0:(num-1)
                p = [p; peaks( spi(l-i,2), 1 ), peaks( spi(l-i,2), 2 ) ];

reply via email to

[Prev in Thread] Current Thread [Next in Thread]