def synthesize(self, method=0, forceReSynthesis=True):
""" function that will synthesise the approximant using the list of atoms
this is mostly DEPRECATED"""
if self.original_signal == None:
_Logger.warning("No original Signal provided")
# return None
if (self.recomposed_signal == None) | forceReSynthesis:
synthesizedSignal = np.zeros(self.length)
if len(self.atoms) == 0:
_Logger.info("No Atoms")
return None
# first method by inverse MDCT
if method == 0:
for mdctSize in self.dico.sizes:
mdctVec = np.zeros(self.length)
for atom in self.atoms:
if atom.length == mdctSize:
# bugFIx
n = atom.time_position + 1
frame = math.floor(
float(n) / float(atom.length / 2)) + 1
# mdctVec[frame*float(atom.length /2) + atom.frequencyBin] += atom.amplitude
mdctVec[frame * float(atom.length /
2) + atom.freq_bin] += atom.mdct_value
synthesizedSignal += imdct(mdctVec, mdctSize)
# synthesizedSignal += concatenate((zeros(mdctSize/4) , imdct(mdctVec ,
# mdctSize)) )[1:-mdctSize/4+1]
# second method by recursive atom synthesis - NOT WORKING
elif method == 1:
for atom in self.atoms:
atom.synthesize_ifft()
synthesizedSignal[atom.time_position:
atom.time_position + atom.length] += atom.waveform
# HACK here to resynthesize using LOMP atoms
elif method == 2:
for atom in self.atoms:
atom.waveForm = atom.synthesize_ifft()
if (atom.proj_score is not None):
if (atom.proj_score < 0):
atom.waveform = (-math.sqrt(-atom.proj_score /
sum(atom.waveform ** 2))) * atom.waveform
else:
atom.waveform = (math.sqrt(atom.proj_score /
sum(atom.waveform ** 2))) * atom.waveform
synthesizedSignal[atom.time_position:
atom.time_position + atom.length] += atom.waveform
self.recomposed_signal = signals.Signal(synthesizedSignal, self.fs)
# return self.recomposed_signal
# other case: we just give the existing synthesized Signal.
return self.recomposed_signal
import numpy as np
from PyMP import Signal, mp
from PyMP.mdct.dico import Dico, LODico
from PyMP.mdct.atom import Atom
print "Running MP, OMP and local versions on synthetic k-sparse"
scales = [16, 64, 256]
dico = Dico(scales)
M = len(scales)
L = 256 * 4
k = 0.2*L
# create a k-sparse signal
sp_vec = np.zeros(M*L,)
from PyMP.tools import mdct
random_indexes = np.arange(M*L)
np.random.shuffle(random_indexes)
random_weights = np.random.randn(M*L)
sp_vec[random_indexes[0:k]] = random_weights[0:k]
sparse_data = np.zeros(L,)
for m in range(M):
sparse_data += mdct.imdct(sp_vec[m*L:(m+1)*L], scales[m])
signal_original = Signal(sparse_data, Fs=8000, mono=True, normalize=False)
signal_original.data += 0.01 * np.random.random(signal_original.length,)
n_atoms = k
signal_original.pad(dico.get_pad())
app_2, dec2 = mp.greedy(signal_original, dico, 100,
n_atoms, debug=0, pad=False, update='locgp')
app_1, dec1 = mp.greedy(signal_original, dico, 100,
n_atoms, debug=0, pad=False, update='mp')
app_3, dec3 = mp.greedy(signal_original, dico, 100,
n_atoms, debug=0, pad=False, update='locomp')