def cwt_iter(fseq,
frange,
nfreqs = 128,
wavelet = pycwt.Morlet(),
normL = None,
max_pixels = None,
cwtfn = pycwt.eds,
verbose = False,
**kwargs):
"""
Iterate over cwt of the time series for each pixel
Parameters:
- `fseq` -- frame sequence instance
- `frange` -- frequency range as a pair or vector of frequencies
- `nfreqs` -- number of frequencies/scales for decomposition
- `wavelet` -- wavelet object [pycwt.Morlet()]
- `normL` -- length of normalizing part (baseline) of the time series
- `max_pixels` -- upper limit on number of pixels to iterate over
- `cwt_fn` -- function to process wavelet coefficients [pycwt.eds]
- `verbose` -- be verbose
- `**kwargs` -- are passed to fseq.pix_iter
Returns:
- generator over (cwt-derived measure, i, j) tuples
(where i,j are frame indices)
"""
tick = time.clock()
L = len(fseq)
subframe = kwargs.has_key('sliceobj') and kwargs['sliceobj'] or None
shape = fseq.shape(subframe)
npix = shape[0]*shape[1]
normL = ifnot(normL, L)
pixel_iter = fseq.pix_iter(**kwargs)
max_pixels = ifnot(max_pixels, npix)
if len(frange) == 2: # a low-high pair
freqs = np.linspace(frange[0], frange[1], num=nfreqs)
else:
freqs= np.array(frange.copy())
nfreqs = len(freqs)
pixel_counter = 0
npix = min(npix, max_pixels)
cwtf = pycwt.cwt_f
dt, tunits = fseq.meta['axes'][0]
for s,i,j in pixel_iter:
# todo: normalization should be optional or as an argument to pix_iter
s = (s-np.mean(s[:normL]))/np.std(s[:normL])
eds = pycwt.eds(cwtf(s, freqs, 1./dt, wavelet, 'zpd'))
pixel_counter+= 1
if verbose:
sys.stderr.write("\rpixel %05d of %05d"%(pixel_counter,npix))
yield eds, i, j
if pixel_counter > max_pixels:
break
if verbose:
sys.stderr.write("\n Finished in %3.2f s\n"%(time.clock()-tick))
def wavelet(nni=None,
rpeaks=None,
fs=4.,
fbands=None,
detrend=True,
show=True,
show_param=True,
legend=True):
# Check input
nn = tools.check_input(nni, rpeaks)
# Verify or set default frequency bands
fbands = fd._check_freq_bands(fbands)
# 外れ値をスプライン補間
nni = _artefact_correction(nni, threshold=0.25)
# 4Hzでリサンプリング
nn_interpol = detrending.resample_to_4Hz(nni, fs)
# 平滑化
if detrend:
nn_interpol = detrending.detrend(nn_interpol, Lambda=500)
freqs = np.arange(0.001, 1, 0.025)
omega0 = 8
r = pycwt.cwt_f(nn_interpol, freqs, fs, pycwt.Morlet(omega0))
rr = np.abs(r)
t_interpol = np.arange(0, len(nn_interpol) / fs, 1. / fs)
plt.imshow(rr,
aspect='auto',
extent=(t_interpol[0], t_interpol[-1], freqs[0], freqs[-1]))
plt.xlabel('Time[s]')
plt.ylabel('Frequency[Hz]')
plt.show()
pass
starttime = float(sys.argv[1])
endtime = float(sys.argv[2])
start = int(starttime * fs)
end = int(endtime * fs)
del starttime, endtime
specdatab = np.array(df[start:end])
specdataa = specdatab.flatten()
fp.close
P = 1024
widths = np.arange(1, P)
cwtmatr = pycwt.cwt_f(specdataa, widths, fs, pycwt.Morlet())
rr = np.abs(cwtmatr)
del cwtmatr
spec = pd.DataFrame(rr)
del specdataa
del rr
s = 100
print(len(df[start:end]))
spec_new = pd.DataFrame(index=spec.index, columns=[])
for i in range(0, len(df[start:end]), s):
spec1 = spec[i]
spec_new[i / s] = spec1
spec_new.T.to_csv(sys.argv[3] + '-' + sys.argv[4] + '-' + sys.argv[1] + '-' +
sys.argv[2] + '-cut.csv',
from swan import pycwt
import numpy as np
import matplotlib.pyplot as plt
x = np.arange(0, 20, 0.01)
y = np.sin(2 * np.pi * 2 * x) * 2 + np.sin(2 * np.pi * 5 * x) * \
2 + np.sin(2 * np.pi * 10 * x)
plt.plot(x, y)
plt.show()
Fs = 1000
omega0 = 8
# (1) Freqを指定してcwt
freqs = np.arange(0.1, 20, 0.025)
r = pycwt.cwt_f(y, freqs, Fs, pycwt.Morlet(omega0))
rr = np.abs(r)
plt.rcParams['figure.figsize'] = (10, 6)
fig = plt.figure()
ax1 = fig.add_axes([0.1, 0.75, 0.7, 0.2])
ax2 = fig.add_axes([0.1, 0.1, 0.7, 0.60], sharex=ax1)
ax3 = fig.add_axes([0.83, 0.1, 0.03, 0.6])
ax1.plot(x, y, 'k')
# , interpolation='nearest')
img = ax2.imshow(np.flipud(rr), extent=[0, 20, 0.1, 20], aspect='auto')
twin_ax = ax2
twin_ax.set_yscale('log')
twin_ax.set_xlim(0, 20)
twin_ax.set_ylim(0.1, 20)
ax2.tick_params(which='both', labelleft=False, left=False)
twin_ax.tick_params(which='both', labelleft=True, left=True, labelright=False)
fig.colorbar(img, cax=ax3)
del datatime
specdatab = np.array(df[start:end])
del df
specdataa = specdatab.flatten()
fp.close
del specdatab
#widths = np.arange(1, 31)
P = 1024
#P = 16384
widths = np.arange(1, P)
plt.subplot(2, 1, 2)
fc = 1.5
cwtmatr = pycwt.cwt_f(specdataa, widths, 25000, pycwt.Morlet(fc))
rr = np.abs(cwtmatr)
del widths
#im = plt.imshow(np.flipud(rr), extent=[starttime, endtime, P, 1], aspect='auto',interpolation='nearest')
im = plt.imshow(np.flipud(rr),
extent=[0, (endtime - starttime) * 1000, P, 1],
aspect='auto',
interpolation='nearest')
#xlim(starttime, endtime)
#xlim(0, (endtime-starttime)*1000)
#ylim(0, 4096)
plt.tick_params(length=10)
xlabel("time [ms]")
ylabel("frequency [Hz] ")
#plt.yscale("log")