Add Janko Slavic findpeaks function

README.md

@@ -9,6 +9,7 @@ This is an overview of all the ready-to-use algorithms I've found to perform pea
 | [peakutils.peak.indexes](#peakutilspeakindexes) | PyPI package PeakUtils<br> Depends on Scipy | Amplitude threshold<br>Minimum distance | ✔ |
 | [peakdetect](#peakdetect-from-sixtenbe) | Single file source<br>Depends on Scipy | Minimum peak distance | ✘ |
 | [Octave-Forge findpeaks](#octave-forge-findpeaks) | Requires an Octave-Forge distribution<br>+ PyPI package oct2py<br>Depends on Scipy | Minimum distance<br>Minimum height<br>Minimum peak width | ✘ |
+| [Janko Slavic findpeaks](#Janko-Slavic-Findpeaks) | Single function<br>Depends on Numpy | Minimum distance<br>Minimum height | ✘ |
 | [Lightweight standalone peaks](#lightweight-standalone-peaks) | Single function<br>Depends on Numpy | Amplitude threshold | ✘ |
 
 ## How to make your choice?
@@ -140,6 +141,26 @@ Requires a rather complicated and not very efficient setup to be called from Pyt
 
 Although the function have an interface close to the MatLab `findpeaks`, it is harder to have the exact same results that with [detect_peaks](#detect_peaks-from-marcos-duarte) or [peakutils.peak.indexes](#peakutilspeakindexes).
 
+## Janko Slavic findpeaks
+
+![](/images/janko_slavic_findpeaks.png?raw=true "Janko Slavic findpeaks")
+
+```python
+import numpy as np
+from vector import vector, plot_peaks
+from libs.findpeaks import findpeaks
+indexes = findpeaks(np.array(vector), spacing=, limit=7)
+print('Peaks are: %s' % (indexes))
+```
+
+[Documentation](https://github.com/jankoslavic/py-tools/blob/master/findpeaks/Findpeaks%20example.ipynb).
+[Source](https://github.com/jankoslavic/py-tools/blob/master/findpeaks/findpeaks.py).
+[Sample code](/tests/janko_slavic_findpeaks.py).
+
+Small and fast peak detection algorithm, with minimum distance and height filtering support. Comes as a single function depending only on Numpy.
+
+The minimum distance filter miss fine granularity tuning (you may filter too many or too few peaks).
+
 ## Lightweight standalone peaks
 
 ![](/images/lightweight_standalone_peaks.png?raw=true "Lightweight standalone peaks")
@@ -155,9 +176,10 @@ print('Peaks are: %s' % (indexes))
 
 [Source and documentation](/tests/lightweight_standalone_peaks.py#L7).
 
-Straightforward, simple and lightweight.
+Straightforward, simple and lightweight peak detection algorithm.
+
+Miss minimum peak height filtering.
 
-Definitely less complete filtering and tuning support than MatLab Signal Processing Toolbox `findpeaks`.
 
 ----------------------------------
 

tests/janko_slavic_findpeaks.py

@@ -0,0 +1,17 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+import numpy as np
+from vector import vector, plot_peaks
+from libs.findpeaks import findpeaks
+
+print('Detect peaks without any filters.')
+indexes = findpeaks(np.array(vector), spacing=1)
+print('Peaks are: %s' % (indexes))
+plot_peaks(np.array(vector), np.array(indexes),
+    algorithm='findpeaks from Janko Slavic')
+
+print('Detect peaks with distance and height filters.')
+indexes = findpeaks(np.array(vector), spacing=2, limit=7)
+print('Peaks are: %s' % (indexes))
+plot_peaks(np.array(vector), np.array(indexes), mph=7, mpd=2,
+    algorithm='findpeaks from Janko Slavic')

tests/libs/findpeaks.py

@@ -0,0 +1,52 @@
+""" Searches for peaks in data
+    History:
+         -nov 2015: Janko Slavic, update
+         -mar 2013: [email protected]
+"""
+
+import numpy as np
+
+def findpeaks(data, spacing=10, limit=None):
+    """Finds peaks in `data` which are of `spacing` width and >=`limit`.
+    :param data: values
+    :param spacing: minimum width for single peak
+    :param limit: peaks should have value greater or equal
+    :return:
+    """
+    x = np.zeros(len(data)+2*spacing)
+    x[:spacing] = data[0]
+    x[-spacing:] = data[-1]
+    x[spacing:-spacing] = data
+    peak_candidate = np.zeros(x.size - 2*spacing - 2)
+    peak_candidate[:] = True
+    for s in range(spacing):
+        h_b = x[s:-2 * spacing + s - 2]  # before
+        h_c = x[spacing: -spacing - 2]  # central
+        h_a = x[spacing + s + 1: -spacing + s - 1]  # after
+        peak_candidate = np.logical_and(peak_candidate, np.logical_and(h_c > h_b, h_c > h_a))
+
+    ind = np.argwhere(peak_candidate)
+    ind = ind.reshape(ind.size)
+    if limit is not None:
+        ind = ind[data[ind] > limit]
+    return ind
+
+
+if __name__ == '__main__':
+    import matplotlib.pyplot as plt
+
+    n = 1000
+    m = 20
+    t = np.linspace(0., 1, n)
+    x = np.zeros(n)
+    np.random.seed(0)
+    phase = 2 * np.pi * np.random.random(m)
+    for i in range(m):
+        x += np.sin(phase[i] + 2 * np.pi * t * i)
+
+    peaks = findpeaks(x, spacing=100, limit=4.)
+    plt.plot(t, x)
+    plt.axhline(4, color='r')
+    plt.plot(t[peaks], x[peaks], 'ro')
+    plt.title('Peaks: minimum value 4., minimum width 100 points')
+    plt.show()