Package edu.gsu.cs.dmlab.imageproc.imageparam.util

Class PeakDetection

java.lang.Object

edu.gsu.cs.dmlab.imageproc.imageparam.util.PeakDetection

All Implemented Interfaces:

IPeakDetector

public class PeakDetection
extends Object
implements IPeakDetector

Peak detection is a designed to find dominant peaks of a time series based on the settings provided by the user. For a peak to be considered dominant, three constraints can be set: a threshold on the frequency domain, a minimum peak-to-peak distance, and a maximum number of dominant peaks.
The main task is done in the method findPeaks which initially finds all the peaks (i.e., any data point whose value is larger than both of its previous and next neighbors), sorts them by their height, and then removes those which do not fall into the set constraints.

Author:

Azim Ahmadzdeh, Data Mining Lab, Georgia State University

Constructor Summary

Constructors

Constructor	Description
PeakDetection(int width, double threshold, int n, boolean isPercentile)	Constructor for the class PeakDetection.

Method Summary

Modifier and Type	Method	Description
static List<Integer>	findCandidatePeaks(double[] data)	It finds the position of any peaks on the time series.
List<Integer>	findPeaks(double[] data)	This method finds the peaks of the given time series based on the constraints provided for the class. Important: The returned list is sorted by the height of the peaks, so the indices are not ordered.
static void	makeTreemapUsingCollections(Map<Integer,Double> m, List<Integer> peaks, double[] data)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

PeakDetection

public PeakDetection(int width,
                     double threshold,
                     int n,
                     boolean isPercentile)

Constructor for the class PeakDetection.

Parameters:

width - is the radius of the neighborhood of an accepted peak within which no other peaks is allowed. To relax this condition, set it to 1 or 0. The results would be similar, i.e., no peaks will be removed because of the neighboring constraints.

threshold - on the frequency domain of the time series below which all the peaks will be ignored.
Note: To relax this condition, set it to the minimum values of the given sequence, and not to zero, unless the sequence is guaranteed to have positive values only.

n - the maximum number of peaks to be found, from the highest to lowest. If n=0, all of the detected peaks will be returned.

isPercentile - if true, then the value of threshold is interpreted as the percentile. Then the accepted values are doubles within [0,100]. If false, then the given double value will be used directly as the actual threshold on the frequency domain.

Method Detail

findPeaks

public List<Integer> findPeaks(double[] data)

This method finds the peaks of the given time series based on the constraints provided for the class.
Important: The returned list is sorted by the height of the peaks, so the indices are not ordered.

Specified by:

findPeaks in interface IPeakDetector

Parameters:

data - The input array to find local maxima in.

Returns:

List of index locations of the found local maxima. The list may or may not be sorted by the height of the peaks, depending on the method implements this interface.

findCandidatePeaks

public static List<Integer> findCandidatePeaks(double[] data)

It finds the position of any peaks on the time series.

Parameters:

data -

Returns:

List of peak index locations

makeTreemapUsingCollections

public static void makeTreemapUsingCollections(Map<Integer,Double> m,
                                               List<Integer> peaks,
                                               double[] data)