from timeside.core import implements, interfacedoc
from timeside.analyzer.core import Analyzer
-from timeside.analyzer import Waveform
+from timeside.analyzer.waveform import Waveform
from timeside.api import IAnalyzer
from numpy import spacing
from collections import deque
self.cor[n] = self.cor[n] + oubli*(self.ftm1[n]*self.et[n]-self.cor[n])
knplus1 = 2*self.cor[n]/(self.variance_f[n] + self.variance_b[n])
-
self.et[n+1] = self.et[n]-knplus1*self.ftm1[n]
self.ft[n+1] = self.ftm1[n]-knplus1*self.et[n]
# along with TimeSide. If not, see <http://www.gnu.org/licenses/>.
# Author: Maxime Le Coz <lecoz@irit.fr>
+from __future__ import absolute_import
from timeside.analyzer.utils import segmentFromValues
from timeside.core import Processor, implements, interfacedoc, FixedSizeInputAdapter
self.block_read = 0
self.pitches = []
self.pitch_confidences = []
-
+
@staticmethod
@interfacedoc
def id():
'''
'''
- frameLenModulation = int(self.decisionLen * self.samplerate() / self.blocksize())
+ frameLenModulation = int(self.decisionLen * self.samplerate() / self.blocksize())
epsilon = numpy.spacing(self.pitch_confidences[0])
w = int(self.decisionLen * self.samplerate() /(self.blocksize()*2))
d = self.pitch_confidences[i-w:i+w]
conf_mean= numpy.mean(d)
conf_var = numpy.var(d+epsilon)
- if self.monoLikelihood(conf_mean,conf_var) > self.polyLikelihood(conf_mean,conf_var) :
+ if self.monoLikelihood(conf_mean,conf_var) > self.polyLikelihood(conf_mean,conf_var) :
is_mono += [True]
else :
is_mono += [False]
-
- conf = self.new_result(data_mode='value', time_mode='framewise')
+
+ conf = self.new_result(data_mode='value', time_mode='framewise')
conf = self.new_result(data_mode='value', time_mode='framewise')
conf.id_metadata.id += '.' + 'yin_confidence'
conf.id_metadata.name += ' ' + 'Yin Confidence'
conf.data_object.value = self.pitch_confidences
-
+
self.process_pipe.results.add(conf)
-
+
convert = {False: 0, True: 1}
label = {0: 'Poly', 1: 'Mono'}
segList = segmentFromValues(is_mono)
segs.label_metadata.label = label
-
+
segs.data_object.label = [convert[s[2]] for s in segList]
segs.data_object.time = [(float(s[0]) * self.blocksize() /
self.samplerate())
self.samplerate())
for s in segList]
self.process_pipe.results.add(segs)
-
+
def monoLikelihood(self,m,v):
-
+
theta1=0.1007;
theta2=0.0029;
beta1=0.5955;
beta2=0.2821;
delta=0.848;
return self.weibullLikelihood(m,v,theta1,theta2,beta1,beta2,delta)
-
-
+
+
def polyLikelihood(self,m,v):
theta1=0.3224;
theta2=0.0121;
beta2=0.8705;
delta=0.644;
return self.weibullLikelihood(m,v,theta1,theta2,beta1,beta2,delta)
-
+
def weibullLikelihood(self,m,v,theta1,theta2,beta1,beta2,delta):
m = numpy.array(m)
v= numpy.array(v)
-
+
c0=numpy.log(beta1*beta2/(theta1*theta2));
a1=m/theta1;
b1=a1**(beta1/delta);
c2=numpy.log(a2);
somme1=(b1+b2)**delta;
Pxy=c0+(beta1/delta-1)*c1+(beta2/delta-1)*c2+(delta-2)*numpy.log(b1+b2)+numpy.log(somme1+1/delta-1)-somme1;
-
+
return numpy.mean(Pxy)
from timeside.analyzer.core import Analyzer
from timeside.analyzer.utils import melFilterBank, computeModulation
from timeside.analyzer.utils import segmentFromValues
-from timeside.analyzer import IRITDiverg
+from timeside.analyzer.irit_diverg import IRITDiverg
from timeside.api import IAnalyzer
from numpy import logical_and,array, hamming, dot, mean, float, arange, nonzero
from numpy.fft import rfft
from timeside.analyzer.core import Analyzer
from timeside.analyzer.utils import melFilterBank, computeModulation
from timeside.analyzer.utils import segmentFromValues
-from timeside.analyzer import IRITDiverg
+from timeside.analyzer.irit_diverg import IRITDiverg
from timeside.api import IAnalyzer
from numpy import mean, diff, arange
from timeside.analyzer.preprocessors import frames_adapter
def __init__(self, blocksize=None, stepsize=None) :
super(IRITMusicSLN, self).__init__();
-
+
self.parents.append(IRITDiverg())
self.wLen = 1.0
self.wStep = 0.1
super(IRITMusicSLN, self).setup(
channels, samplerate, blocksize, totalframes)
self.input_blocksize = int(self.wLen * samplerate)
- self.input_stepsize = int(self.wStep * samplerate)
+ self.input_stepsize = int(self.wStep * samplerate)
@staticmethod
@interfacedoc
def __str__(self):
return "Music confidence indexes"
-
+
@frames_adapter
def process(self, frames, eod=False):
return frames,eod
-
+
def post_process(self):
'''
'''
segList = self.process_pipe.results['irit_diverg.segments'].time
-
+
w = self.wLen/ 2
end = segList[-1]
-
+
tLine = arange(w,end-w,self.wStep)
-
+
# Les plus petits ! <> article
segLen = [mean(diff(getBoundariesInInterval(t-w, t+w, segList))) for t in tLine]
from timeside.analyzer.core import Analyzer
from timeside.analyzer.utils import melFilterBank, computeModulation
from timeside.analyzer.utils import segmentFromValues
-from timeside.analyzer import IRITDiverg
+from timeside.analyzer.irit_diverg import IRITDiverg
from timeside.api import IAnalyzer
from numpy import array, mean, arange, nonzero
from timeside.analyzer.preprocessors import frames_adapter
class IRITMusicSNB(Analyzer):
-
+
implements(IAnalyzer)
def __init__(self, blocksize=1024, stepsize=None, samplerate=None) :
self.wLen = 1.0
self.wStep = 0.1
self.input_blocksize = 0;
- self.input_stepsize = 0;
+ self.input_stepsize = 0;
self.threshold = 20
@interfacedoc
super(IRITMusicSNB, self).setup(
channels, samplerate, blocksize, totalframes)
self.input_blocksize = int(self.wLen * samplerate)
- self.input_stepsize = int(self.wStep * samplerate)
-
+ self.input_stepsize = int(self.wStep * samplerate)
+
@staticmethod
@interfacedoc
def id():
def __str__(self):
return "Music confidence indexes"
-
+
@frames_adapter
def process(self, frames, eod=False):
return frames,eod
-
+
def post_process(self):
'''
'''
-
- segList = self.process_pipe.results['irit_diverg.segments'].time
+
+ segList = self.process_pipe.results['irit_diverg.segments'].time
w = self.wLen/ 2
end = segList[-1]
tLine = arange(0, end, self.wStep)
-
+
segNB = [ len(getBoundariesInInterval(t-w,t+w,segList)) for t in tLine ]
-
+
# Confidence Index
conf = [float(v - self.threshold) / float(self.threshold) if v < 2*self.threshold else 1.0 for v in segNB]
segLenRes = self.new_result(data_mode='value', time_mode='framewise')
# You should have received a copy of the GNU General Public License
# along with TimeSide. If not, see <http://www.gnu.org/licenses/>.
# Author: Maxime Le Coz <lecoz@irit.fr>
+from __future__ import absolute_import
+
from timeside.analyzer.utils import segmentFromValues
from timeside.core import Processor, implements, interfacedoc, FixedSizeInputAdapter
import numpy
from scipy.signal import firwin,lfilter
from scipy.ndimage.morphology import binary_opening,binary_closing
-import pylab
+import pylab
+
-
class IRITStartSeg(Analyzer):
@interfacedoc
def setup(self, channels=None, samplerate=None,
blocksize=None, totalframes=None):
-
+
super(IRITStartSeg, self).setup(channels,
samplerate,
blocksize,
totalframes)
lowFreq = 100.0
-
+
self.input_blocksize = int(0.02 * samplerate)
self.input_stepsize = int(0.008 * samplerate)
-
-
+
+
sr = float(samplerate)
- highFreq = sr/2
+ highFreq = sr/2
f1= lowFreq/sr
- f2= highFreq/sr
- self.filtre = firwin(10, [f1,f2], pass_zero=False)
+ f2= highFreq/sr
+ self.filtre = firwin(10, [f1,f2], pass_zero=False)
self.energy = []
self.maxenergy = 0.002
self.min_overlap = 20
@staticmethod
@interfacedoc
def id():
- return "irit_monopoly"
+ return "irit_startseg"
@staticmethod
@interfacedoc
def __str__(self):
return "Labeled Monophonic/Polyphonic segments"
-
+
@frames_adapter
def process(self, frames, eod=False):
'''
'''
-
+
self.energy += [numpy.sqrt(numpy.mean(lfilter(self.filtre,1.0,frames.T[0])**2))]
return frames, eod
'''
'''
- self.energy = numpy.array(self.energy)/max(self.energy)
+ self.energy = numpy.array(self.energy)/max(self.energy)
silences = numpy.zeros((1,len(self.energy)))[0]
- silences[self.energy<self.maxenergy] = 1
-
+ silences[self.energy<self.maxenergy] = 1
+
step = float(self.input_stepsize) / float(self.samplerate())
- tL = numpy.arange(len(silences))*step
-
+ tL = numpy.arange(len(silences))*step
+
prototype = numpy.load('timeside/analyzer/protoStart2.dat')
prototype2 = numpy.load('timeside/analyzer/protoStart3.dat')
# Lissage pour éliminer les petits segments dans un sens ou l'autre
silencesList.append(tuple(seg))
seg = [i,-1,v]
seg[1] = i
- silencesList.append(tuple(seg))
+ silencesList.append(tuple(seg))
segsList = []
candidates = []
l = len(prototype)
#import pylab
- for s in silencesList :
+ for s in silencesList :
if s[2] == 1 :
shape = numpy.array(self.energy[s[0]:s[1]])
#shape = shape/numpy.max(shape)
d1,_ = computeDist2(prototype,shape)
d2,_ = computeDist2(prototype2,shape)
dist = min([d1,d2])
-
+
candidates.append((s[0],s[1],dist))
#pylab.plot(shape)
#pylab.plot(range(decal,decal+l),prototype)
#pylab.show()
- if dist < self.threshold :
+ if dist < self.threshold :
segsList.append(s)
-
+
label = {0: 'Start',1:'Session'}
- with open('out.lab','w') as f :
- for s in segsList :
+ with open('out.lab','w') as f :
+ for s in segsList :
f.write('%.2f\t%.2f\t%s\n'%(s[0]*step,s[1]*step,label[s[2]]))
-
- with open('cand.lab','w') as f :
- for s in candidates :
- f.write('%.2f\t%.2f\t%f\n'%(s[0]*step,s[1]*step,s[2]))
-
+
+ with open('cand.lab','w') as f :
+ for s in candidates :
+ f.write('%.2f\t%.2f\t%f\n'%(s[0]*step,s[1]*step,s[2]))
+
segs = self.new_result(data_mode='label', time_mode='segment')
segs.id_metadata.id += '.' + 'segments'
segs.id_metadata.name += ' ' + 'Segments'
self.process_pipe.results.add(segs)
-def computeDist2(proto,serie) :
+def computeDist2(proto,serie) :
l = len(proto)
r= range(len(serie))
serie = numpy.array(list(serie)+[0]*(l-1))
v = [numpy.mean(numpy.abs((serie[i:i+l]/numpy.max(serie[i:i+l]))-proto))for i in r]
return numpy.min(v),numpy.argmin(v)
-
+
def computeDist(v1,v2,min_overlap):
'''
-
+
'''
m1 = numpy.argmax(v1)
m2 = numpy.argmax(v2)
l1 = len(v1)
l2 = len(v2)
decal = numpy.abs(m1-m2)
-
+
if m1 >= m2 :
fin = numpy.min([l1-decal,l2])
if fin-decal > min_overlap:
-
+
v1_out = numpy.array(v1[decal:decal+fin])
v2_out = numpy.array(v2[:fin])
d = numpy.mean(numpy.abs(v1_out-v2_out))
else :
v1_out = [0]
v2_out = [1]
- d = 1
+ d = 1
else :
return computeDist(v2, v1,min_overlap)
-
+
return d,v1_out,v2_out
projects / timeside.git / commitdiff