class DynamicLoudnessIterator():
def __init__(self,
model,
fs,
output = None,
globalLoudnessFeature = None,
loudnessLevelFunction = None,
nInputEars = 1):
self.extractor = DynamicLoudnessExtractor(model, fs, nInputEars, output)
self.outputName = model.getOutputModulesToAggregate()[0]
self.converged = False
if globalLoudnessFeature is None:
self.globalLoudnessFeature = np.mean
else:
self.globalLoudnessFeature = globalLoudnessFeature
if loudnessLevelFunction is None:
self.loudnessLevelFunction = asIs
else:
self.loudnessLevelFunction = loudnessLevelFunction
def extractLoudness(self, signal, gainInDecibels = 0):
signalIn = signal.copy() * 10 ** (gainInDecibels / 20.0)
self.extractor.process(signalIn)
timeSeries = self.extractor.outputDict[self.outputName]
loudness = self.globalLoudnessFeature(timeSeries)
loudnessLevel = self.loudnessLevelFunction(loudness)
return loudnessLevel
def process(self, inputSignal, targetLoudness, tol = 0.1, nIters = 5, alpha = 1.0):
# First check if the target loudness is a signal
if type(targetLoudness) is np.ndarray:
targetLoudness = self.extractLoudness(targetLoudness)
storedGain = 0
self.converged = False
for i in range(nIters):
loudnessLevel = self.extractLoudness(inputSignal, storedGain)
error = targetLoudness - loudnessLevel
print (('Gain: %0.3f, Loudness Level: %0.3f, ' +
'Error: %0.3f') % (storedGain, loudnessLevel, error))
if np.abs(error) < tol:
self.converged = True
break
else:
storedGain += error * alpha
if (i == (nIters-1)):
print "Reached iteration limit, not solved within desired error tolerance."
return storedGain
def __init__(
self,
model,
fs,
output=None,
globalLoudnessFeature=None,
loudnessLevelFunction=None,
nInputEars=1
):
self.extractor = DynamicLoudnessExtractor(
model, fs, nInputEars, output
)
self.outputName = model.getOutputModulesToAggregate()[0]
self.converged = False
if globalLoudnessFeature is None:
self.globalLoudnessFeature = np.mean
else:
self.globalLoudnessFeature = globalLoudnessFeature
if loudnessLevelFunction is None:
self.loudnessLevelFunction = asIs
else:
self.loudnessLevelFunction = loudnessLevelFunction
def __init__(self,
model,
fs,
output=None,
globalLoudnessFeature=None,
loudnessLevelFunction=None,
nInputEars=1):
self.extractor = DynamicLoudnessExtractor(model, fs, nInputEars,
output)
self.outputName = model.getOutputModulesToAggregate()[0]
self.converged = False
if globalLoudnessFeature is None:
self.globalLoudnessFeature = np.mean
else:
self.globalLoudnessFeature = globalLoudnessFeature
if loudnessLevelFunction is None:
self.loudnessLevelFunction = asIs
else:
self.loudnessLevelFunction = loudnessLevelFunction
class DynamicLoudnessIterator():
def __init__(self,
model,
fs,
output=None,
globalLoudnessFeature=None,
loudnessLevelFunction=None,
nInputEars=1):
self.extractor = DynamicLoudnessExtractor(model, fs, nInputEars,
output)
self.outputName = model.getOutputModulesToAggregate()[0]
self.converged = False
if globalLoudnessFeature is None:
self.globalLoudnessFeature = np.mean
else:
self.globalLoudnessFeature = globalLoudnessFeature
if loudnessLevelFunction is None:
self.loudnessLevelFunction = asIs
else:
self.loudnessLevelFunction = loudnessLevelFunction
def extractLoudness(self, signal, gainInDecibels=0):
signalIn = signal.copy() * 10**(gainInDecibels / 20.0)
self.extractor.process(signalIn)
timeSeries = self.extractor.outputDict[self.outputName]
loudness = self.globalLoudnessFeature(timeSeries)
loudnessLevel = self.loudnessLevelFunction(loudness)
return loudnessLevel
def process(self,
inputSignal,
targetLoudness,
tol=0.1,
nIters=5,
alpha=1.0):
# First check if the target loudness is a signal
if type(targetLoudness) is np.ndarray:
targetLoudness = self.extractLoudness(targetLoudness)
storedGain = 0
self.converged = False
for i in range(nIters):
loudnessLevel = self.extractLoudness(inputSignal, storedGain)
error = targetLoudness - loudnessLevel
print(('Gain: %0.3f, Loudness Level: %0.3f, ' + 'Error: %0.3f') %
(storedGain, loudnessLevel, error))
if np.abs(error) < tol:
self.converged = True
break
else:
storedGain += error * alpha
if (i == (nIters - 1)):
print "Reached iteration limit, not solved" \
"within desired error tolerance."
return storedGain