下面是一个基于Groovy回答的实际Java实现。(我知道已经发布了Groovy和Kotlin实现，但对于像我这样只做Java的人来说，弄清楚如何在其他语言和Java之间转换真的很麻烦)。

(结果与他人图表相匹配)

算法实现

import java.util.ArrayList;
import java.util.Collections;
import java.util.HashMap;
import java.util.List;

import org.apache.commons.math3.stat.descriptive.SummaryStatistics;

public class SignalDetector {

    public HashMap<String, List> analyzeDataForSignals(List<Double> data, int lag, Double threshold, Double influence) {

        // init stats instance
        SummaryStatistics stats = new SummaryStatistics();

        // the results (peaks, 1 or -1) of our algorithm
        List<Integer> signals = new ArrayList<Integer>(Collections.nCopies(data.size(), 0));

        // filter out the signals (peaks) from our original list (using influence arg)
        List<Double> filteredData = new ArrayList<Double>(data);

        // the current average of the rolling window
        List<Double> avgFilter = new ArrayList<Double>(Collections.nCopies(data.size(), 0.0d));

        // the current standard deviation of the rolling window
        List<Double> stdFilter = new ArrayList<Double>(Collections.nCopies(data.size(), 0.0d));

        // init avgFilter and stdFilter
        for (int i = 0; i < lag; i++) {
            stats.addValue(data.get(i));
        }
        avgFilter.set(lag - 1, stats.getMean());
        stdFilter.set(lag - 1, Math.sqrt(stats.getPopulationVariance())); // getStandardDeviation() uses sample variance
        stats.clear();

        // loop input starting at end of rolling window
        for (int i = lag; i < data.size(); i++) {

            // if the distance between the current value and average is enough standard deviations (threshold) away
            if (Math.abs((data.get(i) - avgFilter.get(i - 1))) > threshold * stdFilter.get(i - 1)) {

                // this is a signal (i.e. peak), determine if it is a positive or negative signal
                if (data.get(i) > avgFilter.get(i - 1)) {
                    signals.set(i, 1);
                } else {
                    signals.set(i, -1);
                }

                // filter this signal out using influence
                filteredData.set(i, (influence * data.get(i)) + ((1 - influence) * filteredData.get(i - 1)));
            } else {
                // ensure this signal remains a zero
                signals.set(i, 0);
                // ensure this value is not filtered
                filteredData.set(i, data.get(i));
            }

            // update rolling average and deviation
            for (int j = i - lag; j < i; j++) {
                stats.addValue(filteredData.get(j));
            }
            avgFilter.set(i, stats.getMean());
            stdFilter.set(i, Math.sqrt(stats.getPopulationVariance()));
            stats.clear();
        }

        HashMap<String, List> returnMap = new HashMap<String, List>();
        returnMap.put("signals", signals);
        returnMap.put("filteredData", filteredData);
        returnMap.put("avgFilter", avgFilter);
        returnMap.put("stdFilter", stdFilter);

        return returnMap;

    } // end
}

主要方法

import java.text.DecimalFormat;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.HashMap;
import java.util.List;

public class Main {

    public static void main(String[] args) throws Exception {
        DecimalFormat df = new DecimalFormat("#0.000");

        ArrayList<Double> data = new ArrayList<Double>(Arrays.asList(1d, 1d, 1.1d, 1d, 0.9d, 1d, 1d, 1.1d, 1d, 0.9d, 1d,
                1.1d, 1d, 1d, 0.9d, 1d, 1d, 1.1d, 1d, 1d, 1d, 1d, 1.1d, 0.9d, 1d, 1.1d, 1d, 1d, 0.9d, 1d, 1.1d, 1d, 1d,
                1.1d, 1d, 0.8d, 0.9d, 1d, 1.2d, 0.9d, 1d, 1d, 1.1d, 1.2d, 1d, 1.5d, 1d, 3d, 2d, 5d, 3d, 2d, 1d, 1d, 1d,
                0.9d, 1d, 1d, 3d, 2.6d, 4d, 3d, 3.2d, 2d, 1d, 1d, 0.8d, 4d, 4d, 2d, 2.5d, 1d, 1d, 1d));

        SignalDetector signalDetector = new SignalDetector();
        int lag = 30;
        double threshold = 5;
        double influence = 0;

        HashMap<String, List> resultsMap = signalDetector.analyzeDataForSignals(data, lag, threshold, influence);
        // print algorithm params
        System.out.println("lag: " + lag + "\t\tthreshold: " + threshold + "\t\tinfluence: " + influence);

        System.out.println("Data size: " + data.size());
        System.out.println("Signals size: " + resultsMap.get("signals").size());

        // print data
        System.out.print("Data:\t\t");
        for (double d : data) {
            System.out.print(df.format(d) + "\t");
        }
        System.out.println();

        // print signals
        System.out.print("Signals:\t");
        List<Integer> signalsList = resultsMap.get("signals");
        for (int i : signalsList) {
            System.out.print(df.format(i) + "\t");
        }
        System.out.println();

        // print filtered data
        System.out.print("Filtered Data:\t");
        List<Double> filteredDataList = resultsMap.get("filteredData");
        for (double d : filteredDataList) {
            System.out.print(df.format(d) + "\t");
        }
        System.out.println();

        // print running average
        System.out.print("Avg Filter:\t");
        List<Double> avgFilterList = resultsMap.get("avgFilter");
        for (double d : avgFilterList) {
            System.out.print(df.format(d) + "\t");
        }
        System.out.println();

        // print running std
        System.out.print("Std filter:\t");
        List<Double> stdFilterList = resultsMap.get("stdFilter");
        for (double d : stdFilterList) {
            System.out.print(df.format(d) + "\t");
        }
        System.out.println();

        System.out.println();
        for (int i = 0; i < signalsList.size(); i++) {
            if (signalsList.get(i) != 0) {
                System.out.println("Point " + i + " gave signal " + signalsList.get(i));
            }
        }
    }
}

结果

lag: 30     threshold: 5.0      influence: 0.0
Data size: 74
Signals size: 74
Data:           1.000   1.000   1.100   1.000   0.900   1.000   1.000   1.100   1.000   0.900   1.000   1.100   1.000   1.000   0.900   1.000   1.000   1.100   1.000   1.000   1.000   1.000   1.100   0.900   1.000   1.100   1.000   1.000   0.900   1.000   1.100   1.000   1.000   1.100   1.000   0.800   0.900   1.000   1.200   0.900   1.000   1.000   1.100   1.200   1.000   1.500   1.000   3.000   2.000   5.000   3.000   2.000   1.000   1.000   1.000   0.900   1.000   1.000   3.000   2.600   4.000   3.000   3.200   2.000   1.000   1.000   0.800   4.000   4.000   2.000   2.500   1.000   1.000   1.000   
Signals:        0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   1.000   0.000   1.000   1.000   1.000   1.000   1.000   0.000   0.000   0.000   0.000   0.000   0.000   1.000   1.000   1.000   1.000   1.000   1.000   0.000   0.000   0.000   1.000   1.000   1.000   1.000   0.000   0.000   0.000   
Filtered Data:  1.000   1.000   1.100   1.000   0.900   1.000   1.000   1.100   1.000   0.900   1.000   1.100   1.000   1.000   0.900   1.000   1.000   1.100   1.000   1.000   1.000   1.000   1.100   0.900   1.000   1.100   1.000   1.000   0.900   1.000   1.100   1.000   1.000   1.100   1.000   0.800   0.900   1.000   1.200   0.900   1.000   1.000   1.100   1.200   1.000   1.000   1.000   1.000   1.000   1.000   1.000   1.000   1.000   1.000   1.000   0.900   1.000   1.000   1.000   1.000   1.000   1.000   1.000   1.000   1.000   1.000   0.800   0.800   0.800   0.800   0.800   1.000   1.000   1.000   
Avg Filter:     0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   1.003   1.003   1.007   1.007   1.003   1.007   1.010   1.003   1.000   0.997   1.003   1.003   1.003   1.000   1.003   1.010   1.013   1.013   1.013   1.010   1.010   1.010   1.010   1.010   1.007   1.010   1.010   1.003   1.003   1.003   1.007   1.007   1.003   1.003   1.003   1.000   1.000   1.007   1.003   0.997   0.983   0.980   0.973   0.973   0.970   
Std filter:     0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.000   0.060   0.060   0.063   0.063   0.060   0.063   0.060   0.071   0.073   0.071   0.080   0.080   0.080   0.077   0.080   0.087   0.085   0.085   0.085   0.083   0.083   0.083   0.083   0.083   0.081   0.079   0.079   0.080   0.080   0.080   0.077   0.077   0.075   0.075   0.075   0.073   0.073   0.063   0.071   0.080   0.078   0.083   0.089   0.089   0.086   

Point 45 gave signal 1
Point 47 gave signal 1
Point 48 gave signal 1
Point 49 gave signal 1
Point 50 gave signal 1
Point 51 gave signal 1
Point 58 gave signal 1
Point 59 gave signal 1
Point 60 gave signal 1
Point 61 gave signal 1
Point 62 gave signal 1
Point 63 gave signal 1
Point 67 gave signal 1
Point 68 gave signal 1
Point 69 gave signal 1
Point 70 gave signal 1

This problem looks similar to one I encountered in a hybrid/embedded systems course, but that was related to detecting faults when the input from a sensor is noisy. We used a Kalman filter to estimate/predict the hidden state of the system, then used statistical analysis to determine the likelihood that a fault had occurred. We were working with linear systems, but nonlinear variants exist. I remember the approach being surprisingly adaptive, but it required a model of the dynamics of the system.

这是一个修改后的Fortran版本的z-score算法。 它是专门针对频率空间中传递函数的峰值(共振)检测进行修改的(每个更改在代码中都有一个小注释)。

如果在输入向量的下界附近存在共振，则第一个修改会向用户发出警告，该共振由高于某个阈值的标准偏差表示(在本例中为10%)。这仅仅意味着信号不够平坦，不足以使检测正确地初始化滤波器。

第二种修改是只将峰值的最大值添加到已找到的峰值中。这是通过将每个发现的峰值与其(滞后)前辈及其(滞后)后继者的大小进行比较来达到的。

第三个变化是尊重共振峰通常在共振频率周围表现出某种形式的对称性。因此，围绕当前数据点对称地计算平均值和std是很自然的(而不仅仅是针对之前的数据点)。这将导致更好的峰值检测行为。

这些修改的效果是，整个信号必须事先被函数知道，这是共振检测的通常情况(类似于Jean-Paul的Matlab示例，其中数据点是动态生成的，这是行不通的)。

function PeakDetect(y,lag,threshold, influence)
    implicit none
    ! Declaring part
    real, dimension(:), intent(in) :: y
    integer, dimension(size(y)) :: PeakDetect
    real, dimension(size(y)) :: filteredY, avgFilter, stdFilter
    integer :: lag, ii
    real :: threshold, influence

    ! Executing part
    PeakDetect = 0
    filteredY = 0.0
    filteredY(1:lag+1) = y(1:lag+1)
    avgFilter = 0.0
    avgFilter(lag+1) = mean(y(1:2*lag+1))
    stdFilter = 0.0
    stdFilter(lag+1) = std(y(1:2*lag+1))

    if (stdFilter(lag+1)/avgFilter(lag+1)>0.1) then ! If the coefficient of variation exceeds 10%, the signal is too uneven at the start, possibly because of a peak.
        write(unit=*,fmt=1001)
1001        format(1X,'Warning: Peak detection might have failed, as there may be a peak at the edge of the frequency range.',/)
    end if
    do ii = lag+2, size(y)
        if (abs(y(ii) - avgFilter(ii-1)) > threshold * stdFilter(ii-1)) then
            ! Find only the largest outstanding value which is only the one greater than its predecessor and its successor
            if (y(ii) > avgFilter(ii-1) .AND. y(ii) > y(ii-1) .AND. y(ii) > y(ii+1)) then
                PeakDetect(ii) = 1
            end if
            filteredY(ii) = influence * y(ii) + (1 - influence) * filteredY(ii-1)
        else
            filteredY(ii) = y(ii)
        end if
        ! Modified with respect to the original code. Mean and standard deviation are calculted symmetrically around the current point
        avgFilter(ii) = mean(filteredY(ii-lag:ii+lag))
        stdFilter(ii) = std(filteredY(ii-lag:ii+lag))
    end do
end function PeakDetect

real function mean(y)
    !> @brief Calculates the mean of vector y
    implicit none
    ! Declaring part
    real, dimension(:), intent(in) :: y
    integer :: N
    ! Executing part
    N = max(1,size(y))
    mean = sum(y)/N
end function mean

real function std(y)
    !> @brief Calculates the standard deviation of vector y
    implicit none
    ! Declaring part
    real, dimension(:), intent(in) :: y
    integer :: N
    ! Executing part
    N = max(1,size(y))
    std = sqrt((N*dot_product(y,y) - sum(y)**2) / (N*(N-1)))
end function std

对于我的应用程序，算法的工作就像一个魅力!

在Palshikar(2009)中发现了另一个算法:

Palshikar, G.(2009)。时间序列中峰值检测的简单算法。在Proc. 1st Int。高级数据分析，商业分析和智能(卷122)。

论文可以从这里下载。

算法是这样的:

algorithm peak1 // one peak detection algorithms that uses peak function S1 

input T = x1, x2, …, xN, N // input time-series of N points 
input k // window size around the peak 
input h // typically 1 <= h <= 3 
output O // set of peaks detected in T 

begin 
O = empty set // initially empty 

    for (i = 1; i < n; i++) do
        // compute peak function value for each of the N points in T 
        a[i] = S1(k,i,xi,T); 
    end for 

    Compute the mean m' and standard deviation s' of all positive values in array a; 

    for (i = 1; i < n; i++) do // remove local peaks which are “small” in global context 
        if (a[i] > 0 && (a[i] – m') >( h * s')) then O = O + {xi}; 
        end if 
    end for 

    Order peaks in O in terms of increasing index in T 

    // retain only one peak out of any set of peaks within distance k of each other 

    for every adjacent pair of peaks xi and xj in O do 
        if |j – i| <= k then remove the smaller value of {xi, xj} from O 
        end if 
    end for 
end

优势

本文提出了5种不同的峰值检测算法 算法在原始时间序列数据上工作(不需要平滑)

缺点

很难事先确定k和h 峰不能是平的(就像我测试数据中的第三个峰)

例子:

这种z-scores方法在峰值检测方面非常有效，也有助于异常值的去除。异常值对话经常讨论每个点的统计价值和变化数据的伦理。

但是，在来自易出错的串行通信或易出错的传感器的重复错误传感器值的情况下，错误或虚假读数中没有统计值。它们需要被识别并移除。

从视觉上看，错误是显而易见的。下图中的直线显示了需要删除的内容。但是用算法识别和消除错误是相当具有挑战性的。z分数效果很好。

下图是通过串行通信从传感器获得的值。偶尔的串行通信错误，传感器错误或两者都导致重复的，明显错误的数据点。

z-score峰值检测器能够在虚假数据点上发出信号，并生成一个干净的结果数据集，同时保留正确数据的特征:

如果边界值或其他标准取决于未来值，那么唯一的解决方案(没有时间机器，或其他关于未来值的知识)是推迟任何决定，直到有足够的未来值。如果你想要一个高于均值的水平，例如，20点，那么你必须等到你至少有19点才能做出任何峰值决策，否则下一个新点可能会完全超过你19点之前的阈值。

Added: If the statistical distribution of the peak heights could be heavy tailed, instead of Uniform or Gaussian, then you may need to wait until you see several thousand peaks before it starts to become unlikely that a hidden Pareto distribution won't produce a peak many times larger than any you currently have seen before or have in your current plot. Unless you somehow know in advance that the very next point can't be 1e20, it could appear, which after rescaling your plot's Y dimension, would be flat up until that point.