def test_subpixel(fig=None):
np.random.seed(2019)
ims = np.random.rand(40, )**2 + 1e1
smooth_ims = smoothImage(ims)
mpos = peak_local_max(smooth_ims, min_distance=3).flatten()
subpos, subval = subpixelmax(smooth_ims, mpos)
np.testing.assert_array_almost_equal(
subpos,
np.array([34.48945739, 19.7971219, 14.04429215, 7.17665281]),
decimal=8)
np.testing.assert_array_almost_equal(
subval,
np.array([10.75670888, 10.46787185, 10.59470934, 10.70051573]),
decimal=8)
if fig:
plt.figure(fig)
plt.clf()
plt.plot(np.arange(ims.size), ims, '.:r', label='data points')
plt.plot(mpos, ims[mpos], 'om', label='integer maxima')
plt.plot(subpos,
subval,
'.g',
markersize=15,
label='subpixel maxima')
plt.legend(numpoints=1)
plt.close('all')
def estimate_dominant_frequency(signal, sample_rate=1, remove_dc=True, fig=None):
""" Estimate dominant frequency in a signal
Args:
signal (array): Input data
sample_rate (float): Sample rate of the data
remove_dc (bool): If True, then do not estimate the DC component
fig (int or None): Optionally plot the estimated frequency
Returns:
Estimated dominant frequency
"""
w = np.fft.fft(signal)
freqs = np.fft.fftfreq(len(signal), d=1. / sample_rate)
if remove_dc:
w[0] = 0
w[freqs < 0] = 0
dominant_idx = np.argmax(np.abs(w))
dominant_frequency = freqs[dominant_idx]
if 0 < dominant_idx < freqs.size / 2 - 1:
dominant_idx_subpixel, _ = subpixelmax(np.abs(w), [dominant_idx])
dominant_frequency = np.interp(dominant_idx_subpixel, np.arange(freqs.size), freqs)[0]
if fig:
plt.figure(fig)
plt.clf()
plt.plot(freqs, np.abs(w), '.b')
plt.xlabel('Frequency')
plt.ylabel('Abs of fft')
plot_vertical_line(dominant_frequency, label='Dominant frequency')
return dominant_frequency