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
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 run():
from os import path
from acoular import __file__ as bpath, MicGeom, WNoiseGenerator, PointSource,\
Mixer, WriteH5, TimeSamples, PowerSpectra, RectGrid, SteeringVector,\
BeamformerBase, L_p
from pylab import figure, plot, axis, imshow, colorbar, show
# set up the parameters
sfreq = 51200
duration = 1
nsamples = duration * sfreq
micgeofile = path.join(path.split(bpath)[0], 'xml', 'array_64.xml')
h5savefile = 'three_sources.h5'
# generate test data, in real life this would come from an array measurement
mg = MicGeom(from_file=micgeofile)
n1 = WNoiseGenerator(sample_freq=sfreq, numsamples=nsamples, seed=1)
n2 = WNoiseGenerator(sample_freq=sfreq,
numsamples=nsamples,
seed=2,
rms=0.7)
n3 = WNoiseGenerator(sample_freq=sfreq,
numsamples=nsamples,
seed=3,
rms=0.5)
p1 = PointSource(signal=n1, mics=mg, loc=(-0.1, -0.1, 0.3))
p2 = PointSource(signal=n2, mics=mg, loc=(0.15, 0, 0.3))
p3 = PointSource(signal=n3, mics=mg, loc=(0, 0.1, 0.3))
pa = Mixer(source=p1, sources=[p2, p3])
wh5 = WriteH5(source=pa, name=h5savefile)
wh5.save()
# analyze the data and generate map
ts = TimeSamples(name=h5savefile)
ps = PowerSpectra(time_data=ts, block_size=128, window='Hanning')
rg = RectGrid( x_min=-0.2, x_max=0.2, y_min=-0.2, y_max=0.2, z=0.3, \
increment=0.01 )
st = SteeringVector(grid=rg, mics=mg)
bb = BeamformerBase(freq_data=ps, steer=st)
pm = bb.synthetic(8000, 3)
Lm = L_p(pm)
# show map
imshow( Lm.T, origin='lower', vmin=Lm.max()-10, extent=rg.extend(), \
interpolation='bicubic')
colorbar()
# plot microphone geometry
figure(2)
plot(mg.mpos[0], mg.mpos[1], 'o')
axis('equal')
show()
sfreq= 12800
n1 = WNoiseGenerator(sample_freq = sfreq,
numsamples = 10*sfreq,
seed = 1)
m = MicGeom()
m.mpos_tot = array([[0,0,0]])
t = PointSource(signal = n1,
mpos = m,
loc = (1, 0, 1))
f = PowerSpectra(time_data = t,
window = 'Hanning',
overlap = '50%',
block_size = 4096)
###################################################################
### Plotting ###
from pylab import figure,plot,show,xlim,ylim,xscale,xticks,xlabel,ylabel,grid,real
from acoular import L_p
band = 3 # octave: 1 ; 1/3-octave: 3
(f_borders, p, f_center) = barspectrum(real(f.csm[:,0,0]), f.fftfreq(), band)
label_freqs = [str(int(_)) for _ in f_center]
figure(figsize=(20, 6))
g = RectGrid(x_min=-0.6,
x_max=-0.0,
y_min=-0.3,
y_max=0.3,
z=0.68,
increment=0.05)
#===============================================================================
# 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 = 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)
wh5.save()
#definition the different source signal
r = 52.5
nsamples = long(sfreq * 0.3)
n1 = WNoiseGenerator(sample_freq=sfreq, numsamples=nsamples)
s1 = SineGenerator(sample_freq=sfreq, numsamples=nsamples, freq=freq)
s2 = SineGenerator(sample_freq=sfreq, numsamples=nsamples, freq=freq, \
phase=pi)
#define a circular array of 8 microphones
m = MicGeom('array_64_8mic.xml')
m.mpos_tot = array([(r*sin(2*pi*i+pi/8), r*cos(2*pi*i+pi/8), 0) \
for i in linspace(0.0, 1.0, 8, False)]).T
t = MaskedTimeSamples(name=datafile)
f = PowerSpectra(time_data=t, window='Hanning', overlap='50%', block_size=128, \
ind_low=1,ind_high=30)
g = RectGrid(x_min=0 - .2,
x_max=0.3,
y_min=-0.13,
y_max=0.3,
z=0,
increment=0.05)
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')
# generate test data, in real life this would come from an array measurement
mg = MicGeom( from_file=micgeofile )
n1 = WNoiseGenerator( sample_freq=sfreq, numsamples=nsamples, seed=1 )
n2 = WNoiseGenerator( sample_freq=sfreq, numsamples=nsamples, seed=2, rms=0.7 )
n3 = WNoiseGenerator( sample_freq=sfreq, numsamples=nsamples, seed=3, rms=0.5 )
p1 = PointSource( signal=n1, mpos=mg, loc=(-0.1,-0.1,0.3) )
p2 = PointSource( signal=n2, mpos=mg, loc=(0.15,0,0.3) )
p3 = PointSource( signal=n3, mpos=mg, loc=(0,0.1,0.3) )
pa = Mixer( source=p1, sources=[p2,p3] )
wh5 = WriteH5( source=pa, name=h5savefile )
wh5.save()
# analyze the data and generate map
ts = TimeSamples( name=h5savefile )
ps = PowerSpectra( time_data=ts, block_size=128, window='Hanning' )
rg = RectGrid( x_min=-0.2, x_max=0.2, y_min=-0.2, y_max=0.2, z=0.3, \
increment=0.01 )
bb = BeamformerBase( freq_data=ps, grid=rg, mpos=mg )
pm = bb.synthetic( 8000, 3 )
Lm = L_p( pm )
# show map
imshow( Lm.T, origin='lower', vmin=Lm.max()-10, extent=rg.extend(), \
interpolation='bicubic')
colorbar()
# plot microphone geometry
figure(2)
plot(mg.mpos[0],mg.mpos[1],'o')
axis('equal')
m.invalid_channels = invalid
g = RectGrid(x_min=-0.6,
x_max=-0.0,
y_min=-0.3,
y_max=0.3,
z=0.68,
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)
# constants
sfreq = 12800 # sample frequency
band = 3 # octave: 1 ; 1/3-octave: 3 (for plotting)
# set up microphone at (0,0,0)
m = MicGeom()
m.mpos_tot = array([[0, 0, 0]])
# create noise source
n1 = WNoiseGenerator(sample_freq=sfreq, numsamples=10 * sfreq, seed=1)
t = PointSource(signal=n1, mics=m, loc=(1, 0, 1))
# create power spectrum
f = PowerSpectra(time_data=t, window='Hanning', overlap='50%', block_size=4096)
# get spectrum data
spectrum_data = real(f.csm[:, 0,
0]) # get power spectrum from cross-spectral matrix
freqs = f.fftfreq() # FFT frequencies
# use barspectrum from acoular.tools to create third octave plot data
(f_borders, p, f_center) = barspectrum(spectrum_data, freqs, band, bar=True)
(f_borders_, p_, f_center_) = barspectrum(spectrum_data,
freqs,
band,
bar=False)
# create figure with barspectra
figure(figsize=(20, 6))
x_max=0.2,
y_min=-0.2,
y_max=0.2,
z_min=0.1,
z_max=0.36,
increment=0.02)
#===============================================================================
# The following provides the cross spectral matrix and defines the CLEAN-SC beamformer.
# To be really fast, we restrict ourselves to only 10 frequencies
# in the range 2000 - 6000 Hz (5*400 - 15*400)
#===============================================================================
f = PowerSpectra(time_data=pa,
window='Hanning',
overlap='50%',
block_size=128,
ind_low=5,
ind_high=16)
st = SteeringVector(grid=g, mics=m, steer_type='true location')
b = BeamformerCleansc(freq_data=f, steer=st)
#===============================================================================
# Calculate the result for 4 kHz octave band
#===============================================================================
map = b.synthetic(4000, 1)
#===============================================================================
# Display views of setup and result.
# For each view, the values along the repsective axis are summed.
# Note that, while Acoular uses a left-oriented coordinate system,
if __name__ == '__main__':
###################################################################
### Defining noise source ###
from acoular import WNoiseGenerator, PointSource, PowerSpectra, MicGeom
sfreq = 12800
n1 = WNoiseGenerator(sample_freq=sfreq, numsamples=10 * sfreq, seed=1)
m = MicGeom()
m.mpos_tot = array([[0, 0, 0]])
t = PointSource(signal=n1, mpos=m, loc=(1, 0, 1))
f = PowerSpectra(time_data=t,
window='Hanning',
overlap='50%',
block_size=4096)
###################################################################
### Plotting ###
from pylab import figure, plot, show, xlim, ylim, xscale, xticks, xlabel, ylabel, grid, real
from acoular import L_p
band = 3 # octave: 1 ; 1/3-octave: 3
(f_borders, p, f_center) = barspectrum(real(f.csm[:, 0, 0]), f.fftfreq(),
band)
label_freqs = [str(int(_)) for _ in f_center]
figure(figsize=(20, 6))
plot(f_borders, L_p(p))
