""" Generate features based on wavelet transformation
.. note:: Currently there is only a simple wrapper around pywt
http://www.pybytes.com/pywavelets/
So one has to install this package.
Scipy does not seem to give a real support for wavelets.
"""
from pySPACE.missions.nodes.base_node import BaseNode
from pySPACE.resources.data_types.feature_vector import FeatureVector
import warnings
import logging
import numpy
try:
import pywt
pywt_error = False
except:
pywt_error = True
[docs]class PywtWaveletNode(BaseNode):
""" Extract features based on the discrete wavelet transform from pywavelets
The components of the wavelet transform are returned as new features.
.. note::
This node is only a wrapper around the
pywavelet package (http://www.pybytes.com/pywavelets/)
**Parameters**
:wavelet:
Name of the wanted wavelet, including the number of taps
The pywt documentation currently names the following possibilities::
:haar family: haar
:db family: db1, db2, db3, db4, db5, db6, db7, db8, db9, db10,
db11, db12, db13, db14, db15, db16, db17, db18,
db19, db20
:sym family: sym2, sym3, sym4, sym5, sym6, sym7, sym8, sym9,
sym10, sym11, sym12, sym13, sym14, sym15, sym16,
sym17, sym18, sym19, sym20
:coif family: coif1, coif2, coif3, coif4, coif5
:bior family: bior1.1, bior1.3, bior1.5, bior2.2, bior2.4,
bior2.6, bior2.8, bior3.1, bior3.3, bior3.5,
bior3.7, bior3.9, bior4.4, bior5.5, bior6.8
:rbio family: rbio1.1, rbio1.3, rbio1.5, rbio2.2, rbio2.4,
rbio2.6, rbio2.8, rbio3.1, rbio3.3, rbio3.5,
rbio3.7, rbio3.9, rbio4.4, rbio5.5, rbio6.8
:dmey family: dmey
which refer to Haar, Daubechies, Symlets, Coiflets, Biorthogonal,
Reverse biorthogonal and Discrete Meyer wavelets
being used to transform the time series.
(*recommended, default: 'haar'*)
:mode:
One of the pywt extension modes::
['zpd', 'cpd', 'sym', 'ppd', 'sp1', 'per']
which mean zero, constant, symmetric, periodic and smooth padding and
periodization. The latter uses the minimum number
of coefficients in comparison to periodic padding.
(*optional, default: 'sym'*)
**Exemplary Call**
.. code-block:: yaml
-
node : DWT
parameters :
mode : 'cpd'
wavelet: 'haar'
:Input: TimeSeries
:Output: FeatureVector
:requires: pywavelets
:Author: Mario Michael Krell (mario.krell@dfki.de)
:Created: 2012/12/04
"""
[docs] def __init__(self, wavelet = None, mode = 'sym',
*args, **kwargs):
super(PywtWaveletNode, self).__init__(*args, **kwargs)
if not wavelet in pywt.wavelist():
warnings.warn("No or wrong wavelet specified (%s). Using 'haar'."%str(wavelet))
wavelet = "haar"
if not mode in pywt.MODES.modes:
warnings.warn("Wrong mode (%s) specified. Using 'sym'."%mode)
mode = 'sym'
self.set_permanent_attributes(wavelet=wavelet,
mode=mode,
feature_names=None,
channel_names=None)
[docs] def _execute(self, x):
""" Extract the wavelet features from the given data x
The feature names will get an A for approximation and
D for details coefficients.
So on example name is: *Waveletname_Channelname_DetailsIndex*,
where index is the position of the coefficient in the
transformed list. Details is *A* or *D*.
Each channel is processed separately and its name is used in *channelname*.
"""
y = x.view(numpy.ndarray)
new_data=[]
if self.feature_names is None:
feature_names=[]
if self.channel_names is None:
self.channel_names = x.channel_names
for n, channel in enumerate(self.channel_names):
cA, cD = pywt.dwt(y[n], wavelet=self.wavelet, mode=self.mode)
for index, value in enumerate(cA):
new_data.append(value)
if self.feature_names is None:
feature_names.append(self.wavelet+"_"+channel+"_A"+str(index))
for index, value in enumerate(cD):
new_data.append(value)
if self.feature_names is None:
feature_names.append(self.wavelet+"_"+channel+"_D"+str(index))
if self.feature_names is None:
self.feature_names = feature_names
feature_vector =\
FeatureVector(numpy.atleast_2d(new_data).astype(numpy.float64),
self.feature_names)
return feature_vector
_NODE_MAPPING = {"DWT": PywtWaveletNode}
