def joint_coding_distortion(target_sig , ref_app, max_rate,
search_width , threshold=0 , doSubtract=True,
debug=0 , discard = False,
precut=-1, initfftw=True):
""" compute the joint coding distortion e.g. given a fixed maximum rate,
what is the distorsion achieved if coding the target_sig knowing the ref_app
This is limited to a time adaptation of the atoms
"""
tolerances = [2]*len(ref_app.dico.sizes)
# initialize the fftw
if initfftw:
mp._initialize_fftw(ref_app.dico, max_thread_num=1)
# initialize factored approx
factorizedApprox = approx.Approx(ref_app.dico,
[],
target_sig,
ref_app.length,
ref_app.fs,
fast_create=True)
timeShifts = np.zeros(ref_app.atom_number)
atom_idx = 0
rate = 0
residual = target_sig.data.copy()
if debug > 0:
print "starting factorization of " , ref_app.atom_number , " atoms"
while (rate < max_rate) and (atom_idx < ref_app.atom_number):
# Stop only when the target rate is achieved or all atoms have been used
atom = ref_app[atom_idx]
# make a copy of the atom
newAtom = atom.copy()
HalfWidth = (tolerances[ref_app.dico.sizes.index(atom.length)] -1) * atom.length/2;
# Search for best time shift using cross correlation
input1 = residual[atom.time_position - HalfWidth : atom.time_position + atom.length + HalfWidth]
input2 = np.zeros((2*HalfWidth + atom.length))
input2[HalfWidth:-HalfWidth] = atom.waveform
if not (input1.shape == input2.shape):
raise ValueError("This certainly happens because you haven't sufficiently padded your signal")
scoreVec = np.array([0.0]);
newts = project_atom(input1,input2 , scoreVec , atom.length)
score = scoreVec
if debug>0:
print "Score of " , score , " found"
# handle MP atoms #
if newAtom.time_shift is not None:
newAtom.time_shift += newts
newAtom.time_position += newts
else:
newAtom.time_shift = newts
newAtom.time_position += newts
atom.proj_score = atom.mdct_value
if debug>0:
print "Factorizing with new offset: " , newts
if score <0:
newAtom.waveform = -newAtom.waveform
factorizedApprox.add(newAtom)
if debug > 0:
print "SRR Achieved of : " , factorizedApprox.compute_srr()
timeShifts[atom_idx] = newAtom.time_shift
if doSubtract:
residual[newAtom.time_position : newAtom.time_position + newAtom.length ] -= newAtom.waveform
rate += np.log2(abs(newts))+1
if debug:
print "Atom %d - rate of %1.3f"%(atom_idx, rate)
atom_idx +=1
# Use to prune the calculus
if atom_idx>precut and precut>0:
curdisto = factorizedApprox.compute_srr()
if curdisto<0:
# pruning
return curdisto
else:
# stop wasting time afterwards
precut = -1
# cleaning
if initfftw:
mp._clean_fftw()
# calculate achieved SNR :
return factorizedApprox.compute_srr()
def runTest(self):
print "-----Test mp sur multi-echelle MDCT"
mdctDico = [32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384]
tol = [2 for i in mdctDico]
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.Signal(op.join(audio_filepath, "ClocheB.wav"),
normalize=True, mono=True)
pyDico2 = dico.Dico(mdctDico)
pyDico_Lomp = dico.LODico(mdctDico)
residualSignal = pySigOriginal.copy()
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)
input_data = 0.42 * np.random.random((N, 1))
projectionMatrix_real = np.zeros((N, 1))
projectionMatrix_comp = np.zeros((N, 1), complex)
scoreTree = np.zeros((P, 1))
pre_twidVec = np.array(
[exp(n * (-1j) * pi / L) for n in range(L)]).reshape(L, 1)
post_twidVec = np.array([exp((
float(n) + 0.5) * -1j * pi * (L / 2 + 1) / L) for n in range(L / 2)]).reshape(L / 2, 1)
print scoreTree.shape, pre_twidVec.shape, post_twidVec.shape
i = 1
j = 10
takeReal = 1
print "Testing Masked Gabor dictionary projections"
print " ---Testing good call"
input_data = 0.42 * np.random.random((N, 1))
projectionMatrix_real = np.zeros((N, 1))
projectionMatrix_comp = np.zeros((N, 1), complex)
projectionMatrix_comp.real = np.random.random((N, 1))
mask_data = np.ones(projectionMatrix_comp.shape)
maskedindexes = [12,35,20]
mask_data[maskedindexes] = 0;
res = parallelProjections.project_masked_gabor(input_data, scoreTree,
projectionMatrix_comp,
mask_data, i, j, L)
if res is not None:
print "Survived.."
if (np.sum(projectionMatrix_comp[maskedindexes]) != 0):
print".. but these should be zeroes..."
print projectionMatrix_comp[maskedindexes]
print mask_data
else:
print np.sum(projectionMatrix_comp)
print ".. and Succeed"
else:
print "Died!"
print " --- Testing MDCT projections with penalized masking"
print " ---Testing good call"
input_data = 0.42 * np.random.random((N, 1))
projectionMatrix = np.zeros((N, 1))
penprojectionMatrix = np.zeros((N, 1))
# test with an overall penalty
pen_mask = np.ones(projectionMatrix.shape)
lamb = 0.01;
res = parallelProjections.project_penalized_mdct(input_data, scoreTree,
projectionMatrix, penprojectionMatrix,
pen_mask, pre_twidVec, post_twidVec, i, j, L,lamb,0)
print np.max(scoreTree), np.max(np.abs(projectionMatrix))
assert(np.max(scoreTree) == np.max(np.abs(penprojectionMatrix)) )
if res is not None:
print "Survived.."
else:
print "Died!"
#print "Testing Bad call:"
#computeMCLT.project(input_data )
print " ---Testing good call"
parallelProjections.project(input_data, scoreTree,
projectionMatrix_real,
pre_twidVec, post_twidVec, i, j, L, 0)
#if parallelFFT.clean_plans() != 1:
# print "Cleaning Stage Failed"
###
##print projectionMatrix_real
print scoreTree
print "--- OK"
#
#
#if computeMCLT.initialize_plans(np.array([L])) != 1:
# print "Initiliazing Stage Failed"
print "---Testing good call: complex"
res = parallelProjections.project_mclt(input_data, scoreTree,
projectionMatrix_comp,
pre_twidVec, post_twidVec, i, j, L)
print scoreTree
if res is not None:
print "--- Ok"
else:
print "ERROR"
#
print "---Testing good call: complex set"
res = parallelProjections.project_mclt_set(input_data, scoreTree,
projectionMatrix_comp,
pre_twidVec, post_twidVec, i, j, L, 1)
if res is not None:
print "--- Ok"
else:
print "ERRRORRRR"
raise TypeError("ARf")
print "---Testing good call: complex Non Linear set with Median"
# Feed it with numpy matrices
sigNumber = 3
NLinput_data = np.concatenate(
(input_data, 0.42 * np.random.randn(N, sigNumber - 1)), axis=1)
NLinput_data = NLinput_data.T
print NLinput_data.shape
NLprojectionMatrix_comp = np.zeros(NLinput_data.shape)
projResult = np.zeros((N, 1))
res = parallelProjections.project_mclt_NLset(NLinput_data, scoreTree,
NLprojectionMatrix_comp,
projResult,
pre_twidVec, post_twidVec, i, j, L, 0)
A = np.median((NLprojectionMatrix_comp) ** 2, axis=0)
#plt.figure()
#plt.plot(A)
#plt.plot(projResult,'r:')
#plt.draw()
#plt.draw()
assert np.sum((A.reshape(projResult.shape) - projResult) ** 2) == 0
if res is not None:
print "--- Ok", scoreTree
else:
print "ERRRORRRR"
raise TypeError("ARf")
print "---Testing good call: complex Non Linear set with Penalized"
projResult = np.zeros((N, 1))
res = parallelProjections.project_mclt_NLset(NLinput_data, scoreTree,
NLprojectionMatrix_comp,
projResult,
pre_twidVec, post_twidVec, i, j, L, 1)
A = (NLprojectionMatrix_comp) ** 2
B = np.sum(A, axis=0)
for l in range(sigNumber):
for m in range(l + 1, sigNumber):
# print i,j
diff = (abs(NLprojectionMatrix_comp[l, :]) - abs(
NLprojectionMatrix_comp[m, :])) ** 2
# print diff
B[:] += diff
#plt.figure()
#plt.plot(B)
#plt.plot(projResult,'r:')
#plt.draw()
#plt.draw()
assert np.sum((B.reshape(projResult.shape) - projResult) ** 2) < 0.000000000001
if res is not None:
print "--- Ok", scoreTree
else:
print "ERRRORRRR"
raise TypeError("ARf")
print "---Testing good call: complex Non Linear set with Weighted"
projResult = np.zeros((N, 1))
scoreTree = np.zeros((P, 1))
res = parallelProjections.project_mclt_NLset(NLinput_data, scoreTree,
NLprojectionMatrix_comp,
projResult,
pre_twidVec, post_twidVec, i, j, L, 2)
A = abs(NLprojectionMatrix_comp)
flatness = (
np.exp((1.0 / sigNumber) * np.sum(np.log(A), axis=0)) / np.mean(A, axis=0))
#print flatness
B = np.multiply(np.nan_to_num(flatness), np.sum(A ** 2, axis=0))
#plt.figure()
#plt.plot(B)
#plt.plot(np.sum(A**2,axis=0),'g')
#plt.plot(projResult,'r:')
#plt.show()
assert np.sum((B.reshape(projResult.shape) - projResult) ** 2) < 0.000000000001
if res is not None:
print "--- Ok", scoreTree
else:
print "ERRRORRRR"
raise TypeError("ARf")
print "---Testing good call: subprojection"
# TODO pass this in the library
res = parallelProjections.subproject(input_data, scoreTree,
projectionMatrix_real, pre_twidVec, post_twidVec, i, j, L, 0, 4)
if res is not None:
print "--- Ok"
else:
print "ERRRORRRR"
raise TypeError("ARf")
#print "---Testing atom projection"
#scoreVec = np.array([0.0])
#
# input2 = np.concatenate( (np.concatenate((np.zeros(L/2) , Atom.waveform) ) ,
# zeros(self.scale/2) ) )
#
#print parallelFFT.project_atom(input_data , input_data , scoreVec )
#
#print "--- Ok"
print "Cleaning"
if parallelProjections.clean_plans() != 1:
print "Cleaning Stage Failed"
print "---testing atom projection and creation"
scale = 128
k = 14
if parallelProjections.initialize_plans(np.array([scale]), np.array([2])) != 1:
print "Initialization Stage Failed"
Atom_test = atom.Atom(scale, 1, 1200, k, 8000)
#Atom_test.mdct_value = 0.57
Atom_test.synthesize(value=1)
#
ref = Atom_test.waveform.copy()
ts = 45
#
input2 = np.concatenate((np.concatenate(
(np.zeros(scale / 2), Atom_test.waveform)), np.zeros(scale / 2)))
input1 = 0.01 * np.random.randn(2 * scale) + np.concatenate((np.concatenate(
(np.zeros(scale / 2 - ts), Atom_test.waveform)), np.zeros(scale / 2 + ts)))
input3 = np.array(input2)
input4 = np.array(input1)
score = np.array([0.0])
import time
nbIt = 10000
t = time.clock()
for j in range(nbIt):
timeShift = parallelProjections.project_atom(input1, input2, score, scale)
print "C code Took", time.clock() - t
t = time.clock()
for j in range(nbIt):
Xcor = np.correlate(input4, input3, "full")
# maxI = abs(Xcor).argmax()
# max = abs(Xcor).max()
print "Numpy took ", time.clock() - t
#print "See if numpy correlate is efficient"
#print score , abs(Xcor).max()
#print timeShift , abs(Xcor).argmax() - 255
#scoreOld = np.array([0.0])
#timeShift2 = computeMCLT.project_atom(input4,input3 ,scoreOld)
score3 = np.array([0.0])
timeShift3 = parallelProjections.project_atom(input1, input2, score3, scale)
#scoreOld2 = np.array([0.0])
#timeShift4 = computeMCLT.project_atom(input4,input3 ,scoreOld2)
#if not(scoreOld == score):
# print "ERROR: new score calculus isn't consistent with old one"
# print scoreOld , score
# print timeShift , timeShift2
# raise TypeError("ARf")
#print score3 , scoreOld2
#
if ts == -timeShift:
print "---- cross-correlation works!"
else:
print "--- ERROR : cross correlation did not pass!"
print timeShift
raise TypeError("ARf")
print timeShift, score
print sum(
Atom_test.waveform * input1[scale / 2 - ts:scale / 2 - ts + Atom_test.length])
#plt.figure()
# plt.plot(np.concatenate( (np.concatenate((np.zeros(scale/2) , ref ) )
# ,np.zeros(scale/2) ) ))
#plt.plot(input1)
#plt.plot(input2)
#plt.legend(('origAtom','signal','newAtom'))
#plt.show()
#
#k = 0
#wf = parallelProjections.get_atom(scale , k)
#wf_gab = parallelProjections.get_real_gabor_atom(scale , k , 0.45)
#
#gabAtom = pymp_GaborAtom.py_pursuit_GaborAtom(scale, 1, 1, k, 1, 0.45)
#wf_gab_test = gabAtom.synthesize()
#plt.figure()
#plt.plot(wf_gab)
#plt.plot(wf_gab_test , 'r:')
#print sum((wf_gab_test - wf_gab)**2)
#
#print (sum(wf_gab**2)) , (sum(wf**2)) , (sum(wf_gab_test**2))
#
#if sum((wf_gab_test - wf_gab)**2) < 0.0000000001:
# print "--- atom construction test OK"
print "Cleaning"
if parallelProjections.clean_plans() != 1:
print "Cleaning Stage Failed"
raise TypeError("ARf")
