Package lib :: Module zero
[frames] | no frames]

Module zero

source code

This is code that, when given a non-negative time-series corresponding to speech loudness, computes the background noise, which is assumed to be constant. It uses a simple model based on a Gaussian and exponential.

Classes
  ZeroProblem
Functions
 
exp(x) source code
tuple(float, float, float, float)
find_zero(x, debug=None)
Models the loudness as being drawn either from a Gaussian distribution (corresponding to the background noise level) or a positive exponential distribtion (corresponding to speech).
source code
float
loud_zero(x, debug=None)
Returns: An estimate of the background noise level
source code
 
zero_sub(x) source code
 
percep_spec_zero(x) source code
 
diagnostic_plot(x, ex) source code
Variables
  PASSLIM = 80
  __package__ = 'lib'

Imports: math, numpy, fpconst, die, load_mod, Numeric_gpk


Function Details

find_zero(x, debug=None)

source code 

Models the loudness as being drawn either from a Gaussian distribution (corresponding to the background noise level) or a positive exponential distribtion (corresponding to speech). It returns parameters of these distributions.

Parameters:
  • x - a time-series of loudness.
Returns: tuple(float, float, float, float)
four floats: (1) the estimate of the background noise level, (2) the width of the noise Gaussian, (3) the mean speech level, (4) the fraction of time with no speech (just background noise). return (zz, sz, e, Pz)

loud_zero(x, debug=None)

source code 
Parameters:
  • x - a time-series of loudness.
Returns: float
An estimate of the background noise level