def fbeamformers():
bb = BeamformerBase(freq_data=f, steer=st, r_diag=True, cached=False)
be = BeamformerEig(freq_data=f, steer=st, r_diag=True, n=54, cached=False)
#frequency beamformers to test
bbase = BeamformerBase(freq_data=f, steer=st, r_diag=True, cached=False)
bc = BeamformerCapon(freq_data=f, steer=st, cached=False)
beig = BeamformerEig(freq_data=f,
steer=st,
r_diag=True,
n=54,
cached=False)
bm = BeamformerMusic(freq_data=f, steer=st, n=6, cached=False)
bd = BeamformerDamas(beamformer=bb, n_iter=10, cached=False)
bdp = BeamformerDamasPlus(beamformer=bb, n_iter=100, cached=False)
bo = BeamformerOrth(beamformer=be,
eva_list=list(range(38, 54)),
cached=False)
bs = BeamformerCleansc(freq_data=f, steer=st, r_diag=True, cached=False)
bcmf = BeamformerCMF(freq_data=f,
steer=st,
method='LassoLarsBIC',
cached=False)
bl = BeamformerClean(beamformer=bb, n_iter=10, cached=False)
bf = BeamformerFunctional(freq_data=f,
steer=st,
r_diag=False,
gamma=3,
cached=False)
bgib = BeamformerGIB(freq_data=f,
steer=st,
method='LassoLars',
n=2,
cached=False)
return (bbase, bc, beig, bm, bl, bo, bs, bd, bcmf, bf, bdp, bgib)
def do_beamforming(self, mic_data):
""" Beamforming using Acoular """
mg, rg, st = self.get_acoular_essentials()
count = 0
#Divide audio samples as per frame rate (10fps) and do beamforming
for s_time in tqdm(
np.arange(self.start_time, self.end_time, self.interval)):
audio_data = mic_data[:,
int(s_time *
self.sample_rate):int((s_time +
self.interval) *
self.sample_rate)]
audio_data = np.transpose(audio_data)
if audio_data.shape[0] == 0:
continue
#Acoular needs audio input through .h5 file
target_file = self.outfile + '/temp.h5'
if os.path.exists(target_file):
os.remove(target_file)
with h5py.File(target_file, 'w') as data_file:
data_file.create_dataset('time_data', data=audio_data)
data_file['time_data'].attrs.__setitem__(
'sample_freq', self.sample_rate)
#.h5 file has issues with closing. Change 'ulimit' if not working
ts = TimeSamples(name=target_file)
ps = PowerSpectra(time_data=ts,
block_size=128,
window='Hanning',
overlap='50%')
bb = BeamformerEig(freq_data=ps, steer=st)
pm = bb.synthetic(self.freq_query, self.octave_band)
Lm = L_p(pm)
if count == 0:
bf_data = np.zeros(
(Lm.shape[0], Lm.shape[1],
len(
np.arange(self.start_time, self.end_time,
self.interval))))
bf_data[:, :, count] = Lm
else:
bf_data[:, :, count] = Lm
count += 1
# remove temp.h5 file after its finished
os.remove(target_file)
return bf_data, rg
def audio_csv_extraction(filepath, trackname, st, firstframe):
if DETAILEDINFO_LVL >= 3:
print('[Azimuth, Elevation], max_value')
ts = WavSamples(name=filepath,start=firstframe*NUM,stop=(firstframe+1)*NUM)
ps = PowerSpectra(time_data=ts, block_size=512, window='Hanning')
be = BeamformerEig(freq_data=ps, steer=st, r_diag=True, n=3)
# be = BeamformerBase(freq_data=ps, steer=st, r_diag=True)
# bb = BeamformerBase(freq_data=ps, steer=st)
# bo = BeamformerOrth(beamformer=be, eva_list=[3])
# Extraktion von Framenummer (Spalte 0), Geräuschklasse (Spalte 1), Azi (Spalte 3) und Ele (Spalte 4) aus .csv-Datei
csvpath = CSV_DIR+trackname+".csv"
rawdata = loadtxt(open(csvpath, "rb"), dtype="int32", delimiter=",", usecols=(0,1,3,4))
if TRAINING or JUST_1SOURCE_FRAMES:
labeldata = rawdata = rm_2sources_frames(rawdata)
if THRESHOLD_FILTER:
soundpressure = zeros((600,1))
## Sound Pressure Level ##
wavsamples = WavSamples(name = filepath).data[:,3]
for frameindex, _ in enumerate(soundpressure[:,0]):
soundpressure[frameindex,0] = sum(wavsamples[frameindex*NUM:(frameindex+1)*NUM]**2)/NUM
spl = L_p(soundpressure[:,0])
##########################
threshold = threshold_calculator(spl)
## Liste der per Threshold gefilterten Frames ##
active_frames = []
for index, frame in enumerate(spl):
if frame > threshold:
active_frames.append(index)
################################################
# active_frames weiterhin eine Liste, aus der man Elemente
# auch zwischendrin löschen kann
# thrdata = array(active_frames).reshape((len(active_frames),1))
thrdata = []
for ind, frame in enumerate(rawdata[:,0]):
if frame in active_frames:
thrdata.append(rawdata[ind,:])
labeldata = array(thrdata)
if not(TRAINING or THRESHOLD_FILTER):
labeldata = rawdata
return ts, be, labeldata
#===============================================================================
f = PowerSpectra(
time_data=t1,
window='Hanning',
overlap='50%',
block_size=128, #FFT-parameters
ind_low=8,
ind_high=16) #to save computational effort, only
# frequencies with indices 8..15 are used
#===============================================================================
# different beamformers in frequency domain
#===============================================================================
bb = BeamformerBase(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04)
bc = BeamformerCapon(freq_data=f, grid=g, mpos=m, c=346.04, cached=False)
be = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, n=54)
bm = BeamformerMusic(freq_data=f, grid=g, mpos=m, c=346.04, n=6)
bd = BeamformerDamas(beamformer=bb, n_iter=100)
bo = BeamformerOrth(beamformer=be, eva_list=list(range(38, 54)))
bs = BeamformerCleansc(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04)
bcmf = BeamformerCMF(freq_data=f, grid=g, mpos=m, c=346.04, \
method='LassoLarsBIC')
bl = BeamformerClean(beamformer=bb, n_iter=100)
bf = BeamformerFunctional(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, \
gamma=4)
#===============================================================================
# plot result maps for different beamformers in frequency domain
#===============================================================================
figure(1, (10, 6))
i1 = 1 #no of subplot
# (in this case, the speed of sound is set)
#===============================================================================
env = Environment(c=346.04)
# =============================================================================
# a steering vector instance. SteeringVector provides the standard freefield
# sound propagation model in the steering vectors.
# =============================================================================
st = SteeringVector(grid=g, mics=m, env=env)
#===============================================================================
# beamformers in frequency domain
#===============================================================================
bb = BeamformerBase(freq_data=f, steer=st, r_diag=True)
bd = BeamformerDamas(beamformer=bb, n_iter=100)
be = BeamformerEig(freq_data=f, steer=st, r_diag=True, n=54)
bo = BeamformerOrth(beamformer=be, eva_list=list(range(38, 54)))
bs = BeamformerCleansc(freq_data=f, steer=st, r_diag=True)
#===============================================================================
# plot result maps for different beamformers in frequency domain
#===============================================================================
fi = 1 #no of figure
for r_diag in (True, False):
figure(fi, (10, 6))
fi += 1
bb.r_diag = r_diag
be.r_diag = r_diag
bs.r_diag = r_diag
i1 = 1 #no of subplot
for steer in ('true level', 'true location', 'classic', 'inverse'):
increment=0.05)
env = Environment(c=346.04)
st = SteeringVector(grid=g, mics=m, env=env)
f = PowerSpectra(
time_data=t1,
window='Hanning',
overlap='50%',
block_size=128, #FFT-parameters
cached=False) #cached = False
bb = BeamformerBase(freq_data=f, steer=st, r_diag=True, cached=False)
bc = BeamformerCapon(freq_data=f, steer=st, cached=False)
be = BeamformerEig(freq_data=f, steer=st, r_diag=True, n=54, cached=False)
bm = BeamformerMusic(freq_data=f, steer=st, n=6, cached=False)
bd = BeamformerDamas(beamformer=bb, n_iter=100, cached=False)
bdp = BeamformerDamasPlus(beamformer=bb, n_iter=100, cached=False)
bo = BeamformerOrth(beamformer=be, eva_list=list(range(38, 54)), cached=False)
bs = BeamformerCleansc(freq_data=f, steer=st, r_diag=True, cached=False)
bcmf = BeamformerCMF(freq_data=f,
steer=st,
method='LassoLarsBIC',
cached=False)
bl = BeamformerClean(beamformer=bb, n_iter=100, cached=False)
bf = BeamformerFunctional(freq_data=f,
steer=st,
r_diag=False,
gamma=4,
cached=False)
#===============================================================================
f = EigSpectra(
time_data=t1,
window='Hanning',
overlap='50%',
block_size=128, #FFT-parameters
ind_low=7,
ind_high=15) #to save computational effort, only
# frequencies with index 1-30 are used
#===============================================================================
# beamformers in frequency domain
#===============================================================================
bb = BeamformerBase(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04)
bd = BeamformerDamas(beamformer=bb, n_iter=100)
be = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, n=54)
bo = BeamformerOrth(beamformer=be, eva_list=list(range(38, 54)))
bs = BeamformerCleansc(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04)
#===============================================================================
# plot result maps for different beamformers in frequency domain
#===============================================================================
fi = 1 #no of figure
for r_diag in (True, False):
figure(fi)
fi += 1
bb.r_diag = r_diag
be.r_diag = r_diag
bs.r_diag = r_diag
i1 = 1 #no of subplot
for steer in ('true level', 'true location', 'classic', 'inverse'):
steer='true level')
bca4Full = BeamformerCapon(freq_data=f,
grid=g,
mpos=m,
r_diag=False,
c=346.04,
steer='true location')
Lbca1Full = L_p(bca1Full.synthetic(4000, 1))
Lbca2Full = L_p(bca2Full.synthetic(4000, 1))
Lbca3Full = L_p(bca3Full.synthetic(4000, 1))
Lbca4Full = L_p(bca4Full.synthetic(4000, 1))
be1Rem = BeamformerEig(freq_data=f,
grid=g,
mpos=m,
r_diag=True,
c=346.04,
steer='classic',
n=12)
be2Rem = BeamformerEig(freq_data=f,
grid=g,
mpos=m,
r_diag=True,
c=346.04,
steer='inverse',
n=12)
be3Rem = BeamformerEig(freq_data=f,
grid=g,
mpos=m,
r_diag=True,
c=346.04,
Lbf4Rem = L_p(bf4Rem.synthetic(4000,1)) Lbf1Full = L_p(bf1Full.synthetic(4000,1)) Lbf2Full = L_p(bf2Full.synthetic(4000,1)) Lbf3Full = L_p(bf3Full.synthetic(4000,1)) Lbf4Full = L_p(bf4Full.synthetic(4000,1)) bca1Full = BeamformerCapon(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, steer='classic') bca2Full = BeamformerCapon(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, steer='inverse') bca3Full = BeamformerCapon(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, steer='true level') bca4Full = BeamformerCapon(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, steer='true location') Lbca1Full = L_p(bca1Full.synthetic(4000,1)) Lbca2Full = L_p(bca2Full.synthetic(4000,1)) Lbca3Full = L_p(bca3Full.synthetic(4000,1)) Lbca4Full = L_p(bca4Full.synthetic(4000,1)) be1Rem = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, steer='classic', n=12) be2Rem = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, steer='inverse', n=12) be3Rem = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, steer='true level', n=12) be4Rem = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, steer='true location', n=12) be1Full = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, steer='classic', n=12) be2Full = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, steer='inverse', n=12) be3Full = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, steer='true level', n=12) be4Full = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=False, c=346.04, steer='true location', n=12) Lbe1Rem = L_p(be1Rem.synthetic(4000,1)) Lbe2Rem = L_p(be2Rem.synthetic(4000,1)) Lbe3Rem = L_p(be3Rem.synthetic(4000,1)) Lbe4Rem = L_p(be4Rem.synthetic(4000,1)) Lbe1Full = L_p(be1Full.synthetic(4000,1)) Lbe2Full = L_p(be2Full.synthetic(4000,1)) Lbe3Full = L_p(be3Full.synthetic(4000,1)) Lbe4Full = L_p(be4Full.synthetic(4000,1))
x = rg.gpos[0]
y = rg.gpos[1]
z = rg.gpos[2]
summed_azi_x = 0.0
summed_azi_y = 0.0
summed_ele = 0.0
summed_max_v = 0.0
print('[Azimuth, Elevation], max_value')
st = SteeringVector(grid=rg, mics=mg)
ts = WavSamples(name=name, start=STARTFRAME * NUM, stop=(STARTFRAME + 1) * NUM)
ps = PowerSpectra(time_data=ts, block_size=512, window='Hanning')
# bb = BeamformerBase(freq_data=ps, steer=st)
be = BeamformerEig(freq_data=ps, steer=st, r_diag=True, n=3)
# bo = BeamformerOrth(beamformer=be, eva_list=[3])
##################### DeepLearning-Matrix ##########################
################ Elevation Azimuth Frequenzbänder Frames
DL_Matrix = zeros((NPOINTS_ELE, NPOINTS_AZI, len(FREQBANDS), FRAMES))
# Wenn .npz, dann keine FRAMES, weil die in einzelne .npy-Tabellen?!
# DL_Matrix = zeros((NPOINTS_ELE, NPOINTS_AZI, len(FREQBANDS)))
for frame_index, frame in enumerate(range(STARTFRAME, ENDFRAME)):
print('### FRAME: ', frame - STARTFRAME, ' (', frame, ') ###')
for freq_index, freq in enumerate(FREQBANDS):
print('FREQ =', freq)
ts.start = frame * NUM
ts.stop = (frame + 1) * NUM
# second, the same with CleanSC
#===============================================================================
bs0 = BeamformerCleansc(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, \
steer='true level')
bs1 = BeamformerCleansc(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, \
steer='true location')
bs2 = BeamformerCleansc(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, \
steer='classic')
bs3 = BeamformerCleansc(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, \
steer='inverse')
#===============================================================================
# third, caching is disabled here for BeamformerEig to save cache storage
# (the results are not needed anyway, exept one time for BeamformerOrth)
#===============================================================================
be = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, n=-1, \
cached=False)
bo = BeamformerOrth(beamformer=be, eva_list=range(38, 54))
#===============================================================================
# save all beamformers in an external file
# important: import dump from cPickle !!!
#===============================================================================
all_bf = (bb0, bb1, bb2, bb3, bs0, bs1, bs2, bs3, bo)
fi = open('all_bf.sav', 'w')
dump(all_bf, fi, -1) # uses newest pickle protocol -1 (default = 0)
fi.close()
#===============================================================================
# plot result maps for different beamformers in frequency domain
#===============================================================================
figure(1)
# for frequency domain methods, this provides the cross spectral matrix and its
# eigenvalues and eigenvectors, if only the matrix is needed then class
# PowerSpectra can be used instead
#===============================================================================
f = EigSpectra(time_data=t1,
window='Hanning', overlap='50%', block_size=128, #FFT-parameters
ind_low=7, ind_high=15) #to save computational effort, only
# frequencies with index 1-30 are used
#===============================================================================
# beamformers in frequency domain
#===============================================================================
bb = BeamformerBase(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04)
bd = BeamformerDamas(beamformer=bb, n_iter=100)
be = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, n=54)
bo = BeamformerOrth(beamformer=be, eva_list=list(range(38,54)))
bs = BeamformerCleansc(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04)
bf = BeamformerFunctional(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, gamma = 60)
#===============================================================================
# plot result maps for different beamformers in frequency domain
#===============================================================================
fi = 1 #no of figure
for r_diag in (True,False):
figure(fi)
suptitle('Old Implementation | R_diag=' + str(r_diag))
fi +=1
# eigenvalues and eigenvectors, if only the matrix is needed then class
# PowerSpectra can be used instead
#===============================================================================
f = EigSpectra(
time_data=t1,
window='Hanning',
overlap='50%',
block_size=128, #FFT-parameters
ind_low=4,
ind_high=15) #to save computational effort, only
# frequencies with index 1-30 are used
#===============================================================================
# different beamformers in frequency domain
#===============================================================================
# first, some simple delay and sum beamformers,
# but with different steering vector types
#===============================================================================
bb0 = BeamformerBase(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, \
steer='true level', cached = False) # this is the default
bs0 = BeamformerCleansc(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, \
steer='true level', cached=False, n=200)
be0 = BeamformerEig(freq_data=f, grid=g, mpos=m, r_diag=True, c=346.04, n=-1, \
steer='true level', cached=False)
bo0 = BeamformerOrth(beamformer=be0, eva_list=range(38, 54), cached=False)
from time import time
ti = time()
map = bs0.result[:]
print time() - ti, g.size
3. Für alle eine Skalierung
'''
FRAME = 275
# Use Folder with 1 Audiofile
WAVFILE = listdir(SINGLE_FILE_TEST)[0]
WAVDIR = SINGLE_FILE_TEST + '/' + WAVFILE
mg, rg, st, _, _ = fbeampreparation()
ts = WavSamples(name=WAVDIR,start=FRAME*NUM,stop=(FRAME+1)*NUM)
ps = PowerSpectra(time_data=ts, block_size=512, window='Hanning')
be = BeamformerEig(freq_data=ps, steer=st, r_diag=True, n=3)
csvpath = CSV_DIR+WAVFILE[:-4]+".csv"
csvdata = loadtxt(open(csvpath, "rb"), dtype="int32", delimiter=",", usecols=(0,3,4))
fr_index = where(csvdata[:,0]==FRAME)[0][0]
if csvdata[fr_index+1,0] != FRAME:
print("Im ausgewählten Frame ist nur 1 Quelle aktiv (nicht 2).")
sys.exit()
Src1_AziEle = csvdata[fr_index,1:3]
Src2_AziEle = csvdata[fr_index+1,1:3]
Src1_Matrix = zeros((NPOINTS_ELE, NPOINTS_AZI, len(FREQBANDS)), dtype="float32")
Src2_Matrix = zeros((NPOINTS_ELE, NPOINTS_AZI, len(FREQBANDS)), dtype="float32")
# Opt 1
maxval1 = zeros(len(FREQBANDS))