Find local maxima using scipy.signal

Question

I have a 1D array (A) of 64 integer values. How can I find the 2 local maxima corresponding to the values 56 and 50 (indices 10 and 45, respectively) using the scipy.signal module?

At first I tried importing

from scipy.signal import find_peaks

A
array([  0.,   1.,   3.,   8.,   6.,  16.,  29.,  29.,  47.,  42.,  56.,
    44.,  49.,  40.,  34.,  34.,  26.,  24.,  25.,  21.,  22.,  24.,
    34.,  17.,  17.,  29.,  24.,  26.,  13.,  25.,  16.,  19.,  19.,
    26.,  24.,  26.,  41.,  34.,  24.,  37.,  37.,  39.,  34.,  40.,
    45.,  50.,  28.,  45.,  43.,  46.,  47.,  41.,  30.,  23.,  19.,
    15.,   9.,  10.,   3.,   6.,   4.,   1.,   1.,   0.])

but before I got any further I got the error message that

"AttributeError: 'module' object has no attribute 'find_peaks'",

so then instead I tried importing

from scipy import signal
peakind = signal.find_peaks_cwt(A, widths=32)

but then I get the error message that

"TypeError: 'int' object is not subscriptable".

I still get this error message even if I first do A.astype(np.int64).

Can't I use scipy.signal_find_peaks on an array of integer values?

Answer 1

For peak detection (1d-vectors or 2d-arrays (images)), you can use the library findpeaks. This library contains multiple peak detection methods.

pip install findpeaks

A = [  0.,   1.,   3.,   8.,   6.,  16.,  29.,  29.,  47.,  42.,  56.,
    44.,  49.,  40.,  34.,  34.,  26.,  24.,  25.,  21.,  22.,  24.,
    34.,  17.,  17.,  29.,  24.,  26.,  13.,  25.,  16.,  19.,  19.,
    26.,  24.,  26.,  41.,  34.,  24.,  37.,  37.,  39.,  34.,  40.,
    45.,  50.,  28.,  45.,  43.,  46.,  47.,  41.,  30.,  23.,  19.,
    15.,   9.,  10.,   3.,   6.,   4.,   1.,   1.,   0.]

Topology method

from findpeaks import findpeaks
fp = findpeaks(method="topology", )
results = fp.fit(A)
results['df']

# For every datapoint, it returns whether it is a peak or valley.

   labx   peak  valley    y   x
0      0  False   False  0.0   0
1      1  False    True  1.0   1
2      1  False   False  3.0   2
3      1  False   False  8.0   3
4      1  False   False  6.0   4
..   ...    ...     ...  ...  ..
59     0  False   False  6.0  59
60     3   True   False  4.0  60
61     0  False   False  1.0  61
62     0  False   False  1.0  62
63     0  False   False  0.0  63
'''

fp.plot()

Peakdetect method

fp = findpeaks(method="peakdetect", lookahead=10)
results = fp.fit(A)
results['df']
fp.plot()

caerus method

fp = findpeaks(method="caerus", params={'minperc':10, 'window':10})
results = fp.fit(A)
results['df']
fp.plot()

Disclaimer: I am the author of this library