def runTest(self):
name = "orchestra"
pySig = Signal(op.join(audio_filepath, "Bach_prelude_40s.wav"), mono=True, normalize=True)
pySig.crop(0, 5 * pySig.fs)
pySig.pad(16384)
sigEnergy = np.sum(pySig.data ** 2)
dico = [128, 1024, 8192]
nbAtoms = 200
classicDIco = mdct_dico.Dico(dico)
spreadDico = mdct_dico.SpreadDico(dico, all_scales=True, penalty=0.1, maskSize=10)
approxClassic, decayClassic = mp.mp(pySig, classicDIco, 20, nbAtoms)
approxSpread, decaySpread = mp.mp(pySig, spreadDico, 20, nbAtoms, pad=False)
import matplotlib.pyplot as plt
plt.figure(figsize=(16, 8))
plt.subplot(121)
approxClassic.plot_tf(ylim=[0, 4000])
plt.title("Classic decomposition : 200 atoms 3xMDCT")
plt.subplot(122)
approxSpread.plot_tf(ylim=[0, 4000])
plt.title("Decomposition with TF masking: 200 atoms 3xMDCT")
# plt.savefig(name + '_TestTFMasking.eps')
plt.figure()
plt.plot([10 * np.log10(i / sigEnergy) for i in decayClassic])
plt.plot([10 * np.log10(i / sigEnergy) for i in decaySpread], "r")
plt.legend(("Classic decomposition", "Spreading Atoms"))
plt.ylabel("Residual energy decay(dB)")
plt.xlabel("Iteration")
def runTest(self):
name = "orchestra"
pySig = Signal(op.join(audio_filepath, "glocs.wav"), mono=True, normalize=True)
pySig.crop(0, 5 * pySig.fs)
pySig.pad(16384)
sigEnergy = np.sum(pySig.data ** 2)
dico = [128, 1024, 8192]
nbAtoms = 200
classicDIco = mdct_dico.Dico(dico, useC=False)
spreadDico = mdct_dico.SpreadDico(
dico, all_scales=False, spread_scales=[1024, 8192], penalty=0.1, mask_time=2, mask_freq=2
)
approxClassic, decayClassic = mp.mp(pySig, classicDIco, 20, nbAtoms)
approxSpread, decaySpread = mp.mp(pySig, spreadDico, 20, nbAtoms, pad=False)
plt.figure(figsize=(16, 8))
plt.subplot(121)
approxClassic.plot_tf(ylim=[0, 4000])
plt.title("Classic decomposition : 200 atoms 3xMDCT")
plt.subplot(122)
approxSpread.plot_tf(ylim=[0, 4000])
plt.title("Decomposition with TF masking: 200 atoms 3xMDCT")
# plt.savefig(name + '_TestTFMasking.eps')
plt.figure()
plt.plot([10 * np.log10(i / sigEnergy) for i in decayClassic])
plt.plot([10 * np.log10(i / sigEnergy) for i in decaySpread], "r")
plt.legend(("Classic decomposition", "Spreading Atoms"))
plt.ylabel("Residual energy decay(dB)")
plt.xlabel("Iteration")
# plt.savefig(name + '_decayTFMasking.eps')
plt.figure()
for blockI in range(1, 3):
block = spreadDico.blocks[blockI]
plt.subplot(2, 2, blockI)
print block.mask.shape, block.mask.shape[0] / (block.scale / 2), block.scale / 2
plt.imshow(
np.reshape(block.mask, (block.mask.shape[0] / (block.scale / 2), block.scale / 2)),
interpolation="nearest",
aspect="auto",
)
plt.colorbar()
plt.subplot(2, 2, blockI + 2)
# print block.mask.shape, block.mask.shape[0] / (block.scale/2),
# block.scale/2
block.im_proj_matrix()
plt.colorbar()
def recompute(self, signal=None, **kwargs):
for key in kwargs:
self.params[key] = kwargs[key]
if signal is not None:
if isinstance(signal, str):
# TODO allow for stereo signals
signal = Signal(signal, normalize=True, mono=True)
self.orig_signal = signal
if self.orig_signal is None:
raise ValueError("No original Sound has been given")
if self.params.has_key('fs'):
self.orig_signal.resample(self.params['fs'])
self.params['fs'] = self.orig_signal.fs
mdct_dico = self._get_dico()
from PyMP import mp
self.rep = mp.mp(self.orig_signal,
mdct_dico,
self.params['SRR'],
self.params['n_atoms'],
silent_fail=True,
pad=self.params['pad'],
debug=self.params['debug'])[0]
def runTest(self):
print "------------------ Test3 Populate from MP coeffs ---------"
fileIndex = 2
RandomAudioFilePath = file_names[fileIndex]
print 'Working on %s' % RandomAudioFilePath
sizes = [2**j for j in range(7, 15)]
segDuration = 5
nbAtom = 20
pySig = signals.Signal(op.join(audio_files_path, RandomAudioFilePath),
mono=True,
normalize=True)
segmentLength = ((segDuration * pySig.fs) / sizes[-1]) * sizes[-1]
nbSeg = floor(pySig.length / segmentLength)
# cropping
pySig.crop(0, segmentLength)
# create dictionary
pyDico = Dico(sizes)
approx, decay = mp.mp(pySig, pyDico, 20, nbAtom, pad=True, debug=0)
ppdb = XMDCTBDB('MPdb.db', load=False)
# ppdb.keyformat = None
ppdb.populate(approx, None, fileIndex)
nKeys = ppdb.get_stats()['ndata']
# compare the number of keys in the base to the number of atoms
print ppdb.get_stats()
self.assertEqual(nKeys, approx.atom_number)
# now try to recover the fileIndex knowing one of the atoms
Key = [
log(approx.atoms[0].length, 2),
approx.atoms[0].reduced_frequency * pySig.fs
]
T, fileI = ppdb.get(Key)
Treal = (float(approx.atoms[0].time_position) / float(pySig.fs))
print T, Treal
self.assertEqual(fileI[0], fileIndex)
Tpy = np.array(T)
self.assertTrue((np.abs(Tpy - Treal)).min() < 0.1)
# last check: what does a request for non-existing atom in base return?
T, fileI = ppdb.get((11, 120.0))
self.assertEqual(T, [])
self.assertEqual(fileI, [])
# now let's just retrieve the atoms from the base and see if they are
# the same
histograms = ppdb.retrieve(approx, None, offset=0)
# plt.figure()
# plt.imshow(histograms[0:10,:])
# plt.show()
del ppdb
def runTest(self):
dico = [128, 1024, 8192]
nbAtoms = 100
pySig = Signal(op.join(audio_filepath, "Bach_prelude_40s.wav"), mono=True, normalize=True)
classicDIco = mdct_dico.Dico(dico)
spreadDico = mdct_dico.SpreadDico(dico, all_scales=True, penalty=0, maskSize=10)
import time
t = time.time()
app_mp, _ = mp.mp(pySig, classicDIco, 20, nbAtoms)
print "Classic took %1.3f sec" % (time.time() - t)
t = time.time()
app_spreadmp, _ = mp.mp(pySig, spreadDico, 20, nbAtoms)
print "Spread took %1.3f sec" % (time.time() - t)
plt.figure()
plt.subplot(121)
app_mp.plot_tf()
plt.subplot(122)
app_spreadmp.plot_tf()
plt.show()
def recompute(self, signal=None, **kwargs):
for key in kwargs:
self.params[key] = kwargs[key]
if signal is not None:
if isinstance(signal, str):
# TODO allow for stereo signals
signal = Signal(signal, normalize=True, mono=True)
self.orig_signal = signal
if self.orig_signal is None:
raise ValueError("No original Sound has been given")
if self.params.has_key('fs'):
self.orig_signal.downsample(self.params['fs'])
# print "Downsampling"
if self.params.has_key('crop'):
self.orig_signal.crop(0, self.params['crop'])
# print "Cropping"
if self.params.has_key('pad'):
self.orig_signal.pad(self.params['pad'])
# print "Padding"
self.params['fs'] = self.orig_signal.fs
dico = self._get_dico()
from PyMP import mp
self.rep = mp.mp(self.orig_signal,
dico,
self.params['SRR'],
self.params['n_atoms'],
silent_fail=True,
pad=False,
debug=self.params['debug'],
max_thread_num=3)[0]
def runTest(self):
''' take the base previously constructed and retrieve the song index based on 10 atoms
'''
print "------------------ Test6 recognition ---------"
nbCandidates = 8
ppdb = XMDCTBDB('LargeMPdb.db', load=True)
print 'Large Db of ' + str(ppdb.get_stats()['nkeys']) + ' and ' + str(
ppdb.get_stats()['ndata'])
# Now take a song, decompose it and try to retrieve it
fileIndex = 6
RandomAudioFilePath = file_names[fileIndex]
print 'Working on ' + str(RandomAudioFilePath)
pySig = signals.Signal(op.join(audio_files_path, RandomAudioFilePath),
mono=True)
pyDico = LODico(sizes)
segDuration = 5
offsetDuration = 7
offset = offsetDuration * pySig.fs
nbAtom = 50
segmentLength = ((segDuration * pySig.fs) / sizes[-1]) * sizes[-1]
pySig.crop(offset, offset + segmentLength)
approx, decay = mp.mp(pySig, pyDico, 40, nbAtom, pad=True)
# plt.figure()
# approx.plotTF()
# plt.show()
res = map(ppdb.get, map(ppdb.kform, approx.atoms),
[(a.time_position - pyDico.get_pad()) / approx.fs
for a in approx.atoms])
#
#res = map(bdb.get, map(bdb.kform, approx.atoms))
histogram = np.zeros((600, nbCandidates))
for i in range(approx.atom_number):
print res[i]
histogram[res[i]] += 1
max1 = np.argmax(histogram[:])
Offset1 = max1 / nbCandidates
estFile1 = max1 % nbCandidates
# candidates , offsets = ppdb.retrieve(approx);
# print approx.atom_number
histograms = ppdb.retrieve(approx, None, offset=0, nbCandidates=8)
# print histograms , np.max(histograms) , np.argmax(histograms, axis=0) ,
# np.argmax(histograms, axis=1)
# plt.figure()
# plt.imshow(histograms[0:20,:],interpolation='nearest')
# plt.show()
maxI = np.argmax(histograms[:])
OffsetI = maxI / nbCandidates
estFileI = maxI % nbCandidates
print fileIndex, offsetDuration, estFileI, OffsetI, estFile1, Offset1, max1, maxI
import matplotlib.pyplot as plt
# plt.figure(figsize=(12,6))
# plt.subplot(121)
# plt.imshow(histograms,aspect='auto',interpolation='nearest')
# plt.subplot(122)
# plt.imshow(histogram,aspect='auto',interpolation='nearest')
## plt.imshow(histograms,aspect='auto',interpolation='nearest')
## plt.colorbar()
# plt.show()
print maxI, OffsetI, estFileI
self.assertEqual(histograms[OffsetI, estFileI], np.max(histograms))
self.assertEqual(fileIndex, estFileI)
self.assertTrue(abs(offsetDuration - OffsetI) <= 2.5)
def runTest(self):
ppdb = XMDCTBDB('tempdb.db',
load=False,
persistent=True,
time_max=500.0)
pySig = signals.LongSignal(op.join(audio_files_path, file_names[0]),
frame_duration=5,
mono=False,
Noverlap=0)
self.assertEqual(pySig.segment_size, 5.0 * pySig.fs)
max_nb_seg = 10
nb_atoms = 150
scales = SpreadDico([8192], penalty=0.1, mask_time=2, mask_freq=20)
# scales = Dico([8192])
for segIdx in range(min(max_nb_seg, pySig.n_seg)):
pySigLocal = pySig.get_sub_signal(segIdx,
1,
mono=True,
normalize=False,
channel=0,
pad=scales.get_pad())
print "MP on segment %d" % segIdx
# run the decomposition
approx, decay = mp.mp(pySigLocal, scales, 2, nb_atoms, pad=False)
print "Populating database with offset " + str(
segIdx * pySig.segment_size / pySig.fs)
ppdb.populate(approx,
None,
0,
offset=float((segIdx * pySig.segment_size) -
scales.get_pad()) / float(pySig.fs))
# ok we have a DB with only 1 file and different segments, now
nb_test_seg = 15
long_sig_test = signals.LongSignal(op.join(audio_files_path,
file_names[0]),
frame_duration=5,
mono=False,
Noverlap=0.5)
count = 0
for segIdx in range(min(nb_test_seg, long_sig_test.n_seg)):
pySigLocal = long_sig_test.get_sub_signal(segIdx,
1,
mono=True,
normalize=False,
channel=0,
pad=scales.get_pad())
# print "MP on segment %d" % segIdx
# run the decomposition
approx, decay = mp.mp(pySigLocal, scales, 2, nb_atoms, pad=False)
print approx.atom_number
histograms = ppdb.retrieve(approx, None, nbCandidates=1)
maxI = np.argmax(histograms[:])
OffsetI = maxI / 1
estFileI = maxI % 1
oracle_value = segIdx * long_sig_test.segment_size * (
1 - long_sig_test.overlap) / long_sig_test.fs
print "Seg %d Oracle: %1.1f - found %1.1f" % (segIdx, oracle_value,
OffsetI)
if abs(OffsetI - oracle_value) < 5:
count += 1
glob = float(count) / float(min(nb_test_seg, long_sig_test.n_seg))
print "Global Score of %1.3f" % glob
self.assertGreater(glob, 0.8)
def runTest(self):
''' time to test the fingerprinting scheme, create a base with 10 atoms for 8 songs, then
Construct the histograms and retrieve the fileIndex and time offset that is the most
plausible '''
print "------------------ Test5 DB construction ---------"
# # create the base : persistent
ppdb = XMDCTBDB('LargeMPdb.db', load=False, time_res=0.2)
print ppdb
padZ = 2 * sizes[-1]
# BUGFIX: pour le cas MP classique: certains atome reviennent : pas
# cool car paire key/data existe deja!
pyDico = LODico(sizes)
segDuration = 5
nbAtom = 50
sig = signals.LongSignal(op.join(audio_files_path, file_names[0]),
frame_size=sizes[-1],
mono=False,
Noverlap=0)
segmentLength = ((segDuration * sig.fs) / sizes[-1]) * sizes[-1]
max_seg_num = 5
# " run MP on a number of files"
nbFiles = 8
keycount = 0
for fileIndex in range(nbFiles):
RandomAudioFilePath = file_names[fileIndex]
print fileIndex, RandomAudioFilePath
if not (RandomAudioFilePath[-3:] == 'wav'):
continue
pySig = signals.LongSignal(op.join(audio_files_path,
RandomAudioFilePath),
frame_size=segmentLength,
mono=False,
Noverlap=0)
nbSeg = int(pySig.n_seg)
print 'Working on ' + str(RandomAudioFilePath) + ' with ' + str(
nbSeg) + ' segments'
for segIdx in range(min(nbSeg, max_seg_num)):
pySigLocal = pySig.get_sub_signal(segIdx,
1,
True,
True,
channel=0,
pad=padZ)
print "MP on segment %d" % segIdx
# run the decomposition
approx, decay = mp.mp(pySigLocal,
pyDico,
40,
nbAtom,
pad=False)
print "Populating database with offset " + str(
segIdx * segmentLength / sig.fs)
ppdb.populate(approx,
None,
fileIndex,
offset=float((segIdx * segmentLength) - padZ) /
sig.fs)
keycount += approx.atom_number
print ppdb.get_stats()
import matplotlib.pyplot as plt
import os
from PyMP import Signal, mp, mp_coder
from PyMP.mdct import Dico
abPath = os.path.abspath("../../data/")
sig = Signal(abPath + "/ClocheB.wav", mono=True) # Load Signal
sig.crop(0, 4.0 * sig.fs) # Keep only 4 seconds
# atom of scales 8, 64 and 512 ms
scales = [(s * sig.fs / 1000) for s in (8, 64, 512)]
# Dictionary for Standard MP
pyDico = Dico(scales)
# Launching decomposition, stops either at 20 dB of SRR or 2000 iterations
mpApprox, mpDecay = mp.mp(sig, pyDico, 20, 2000)
# mpApprox.atomNumber
SNR, bitrate, quantizedApprox = mp_coder.simple_mdct_encoding(mpApprox, 2000, Q=14)
quantizedApprox.plot_tf()
plt.show()
abPath = os.path.abspath('../../data/')
sig = Signal(abPath + '/glocs.wav', mono=True, normalize=True)
# taking only the first musical phrase (3.5 seconds approximately)
sig.crop(0, 3.5 * sig.fs)
sig.pad(8192)
# add some minor noise to avoid null areas
sig.data += 0.0001 * np.random.randn(sig.length)
# create MDCT multiscale dictionary
dico = Dico(sizes)
# run the MP routine
approx, decay = mp.mp(sig, dico, 50, n_atoms)
# plotting the results
timeVec = np.arange(0, float(sig.length)) / sig.fs
plt.figure(figsize=(10, 6))
axOrig = plt.axes([0.05, 0.55, .4, .4])
axOrig.plot(timeVec, sig.data)
axOrig.set_title('(a)')
axOrig.set_xticks([1, 2, 3, 4])
axOrig.set_ylim([-1.0, 1.0])
axApprox = plt.axes([0.05, 0.07, .4, .4])
axApprox.plot(timeVec, approx.recomposed_signal.data)
axApprox.set_title('(c)')
axApprox.set_xlabel('Temps (s)')
import numpy as np os.environ['PYMP_PATH'] = '/home/manu/workspace/PyMP/' from PyMP.mdct import Dico from PyMP import mp, Signal from PyMP.tools.Misc import euclid_dist, hamming_dist # Decomposing and visualizing the sparse dec signal = Signal(op.join(os.environ['PYMP_PATH'],'data/Bach_prelude_4s.wav'), mono=True) sig_occ1 = signal[:signal.length/2] sig_occ2 = signal[signal.length/2:] dico = Dico([128,1024,8192]) target_srr = 5 max_atom_num = 200 app_1, _ = mp.mp(sig_occ1, dico, target_srr, max_atom_num) app_2, _ = mp.mp(sig_occ2, dico, target_srr, max_atom_num) #plt.figure(figsize=(16,6)) #plt.subplot(121) #app_1.plot_tf() #plt.subplot(122) #app_2.plot_tf() #plt.show() sp_vec_1 = app_1.to_array()[0] sp_vec_2 = app_2.to_array()[0] print "%1.5f, %1.5f"%(euclid_dist(sp_vec_1,sp_vec_2), hamming_dist(sp_vec_1,sp_vec_2))
print "test the initialization function"
if parallelProjections.initialize_plans(np.array(mdctDico), np.array(tol)) != 1:
print "Initiliazing Stage Failed"
if parallelProjections.clean_plans() != 1:
print "Initiliazing Stage Failed"
pySigOriginal = signals.InitFromFile("../../data/ClocheB.wav", True, True)
pyDico2 = dico.Dico(mdctDico)
pyDico_Lomp = dico.LODico(mdctDico)
residualSignal = pySigOriginal.copy()
app, decay = mp.mp(pySigOriginal, pyDico2, 20, 200, 0)
print " profiling test with C integration"
cProfile.runctx("mp.mp(pySigOriginal, pyDico2, 20, 200 ,0)", globals(), locals())
cProfile.runctx("mp.mp(pySigOriginal, pyDico_Lomp, 20, 200 ,0)", globals(), locals())
################" C binding tests ########
N = 64
L = 16
if parallelProjections.initialize_plans(np.array([L]), np.array([2])) != 1:
print "Initiliazing Stage Failed"
P = N / (L / 2)
mpl.rcParams['legend.fancybox'] = True
mpl.rcParams['legend.shadow'] = True
mpl.rcParams['image.interpolation'] = 'Nearest'
#mpl.rcParams['text.usetex'] = True
# Load glockenspiel signal
abPath = os.path.abspath('../../data/')
sig = Signal(abPath + '/glocs.wav', mono=True, normalize=True)
sig.crop(0, 3 * sig.fs)
scales = [128, 1024, 8192]
n_atoms = 500
srr = 30
mp_dico = Dico(scales)
lomp_dico = LODico(scales)
mp_approx, mp_decay = mp.mp(sig, mp_dico, srr, n_atoms, pad=True)
lomp_approx, lomp_decay = mp.mp(sig, lomp_dico, srr, n_atoms, pad=False)
plt.figure()
plt.subplot(211)
mp_approx.plot_tf()
plt.subplot(212)
lomp_approx.plot_tf()
# print mp_approx , lomp_approx
plt.show()
M. Moussallam
"""
from PyMP.mdct import Dico, LODico
from PyMP.mdct.rand import SequenceDico
from PyMP import mp, mp_coder, Signal
signal = Signal('../data/ClocheB.wav', mono=True) # Load Signal
signal.crop(0, 4.0 * signal.fs) # Keep only 4 seconds
# atom of scales 8, 64 and 512 ms
scales = [(s * signal.fs / 1000) for s in (8, 64, 512)]
signal.pad(scales[-1])
# Dictionary for Standard MP
dico = Dico(scales)
# Launching decomposition, stops either at 20 dB of SRR or 2000 iterations
app, dec = mp.mp(signal, dico, 20, 2000, pad=False)
app.atom_number
snr, bitrate, quantized_app = mp_coder.simple_mdct_encoding(
app, 8000, Q=14)
print (snr, bitrate)
print "With Q=5"
snr, bitrate, quantized_app = mp_coder.simple_mdct_encoding(
app, 8000, Q=5)
print (snr, bitrate)
snr, bitrate, quantized_app = mp_coder.simple_mdct_encoding(
app, 2000, Q=14)
