Gaussian Model¶

class clustering.gauss.GaussFull(name, dim)[source]¶

Class to model a speaker by a gaussian with full covariance

add(features)[source]¶: Accumulate statistics for features :param features: numpy.ndarray

compute()[source]¶

Compute the mean and covariance according the statistique, the log det: of the covariance and the partial BIC $PBIC$ .

$PBIC_{x} = \frac{n_x}{2} \log|\Sigma_x|$

classmethod cst_bic(dim, alpha)[source]¶

Compute the BIC constant:

$cst = \frac{1}{2} \alpha \left(d + \frac{d(d+1)}{2}\right)$

where $d`is the feature dimension (*dim*) and :math:`alpha$ a threshold (alpha)

Parameters:	dim – the feature dimension alpha – the threshold
Returns:	the constant

classmethod merge(m1, m2)[source]¶

Merge two models m1 and m2. Compute the new mean (mu), covariance (cov) and PBIC partial_bic

Parameters:	m1 – a GaussFull object m2 – a GaussFull object
Returns:	a GaussFull object

classmethod merge_partial_bic(m1, m2)[source]¶

Merge statistic accumulators of two a GaussFull objects, compute the new log det.

Parameters:	m1 – a GaussFull object m2 – a GaussFull object
Returns:	the log det