gmm_scoring

Copyright 2014-2019 Anthony Larcher and Sylvain Meignier

features_server provides methods to test gmm models

gmm_scoring.gmm_scoring(ubm, enroll, ndx, feature_server, num_thread=1)[source]

Compute log-likelihood ratios for sequences of acoustic feature frames between a Universal Background Model (UBM) and a list of Gaussian Mixture Models (GMMs) which only mean vectors differ from the UBM.

Parameters

  • ubm – a Mixture object used to compute the denominator of the likelihood ratios

  • enroll – a StatServer object which stat1 attribute contains mean super-vectors of the GMMs to use to compute the numerator of the likelihood ratios.

  • ndx – an Ndx object which define the list of trials to compute

  • feature_server – a FeatureServer object to load the features

  • num_thread – number of thread to launch in parallel

Returns

a Score object.

gmm_scoring.gmm_scoring_singleThread(ubm, enroll, ndx, feature_server, score_mat, seg_idx=None)[source]

Compute log-likelihood ratios for sequences of acoustic feature frames between a Universal Background Model (UBM) and a list of Gaussian Mixture Models (GMMs) which only mean vectors differ from the UBM.

Parameters

  • ubm – a Mixture object used to compute the denominator of the likelihood ratios

  • enroll – a StatServer object which stat1 attribute contains mean super-vectors of the GMMs to use to compute the numerator of the likelihood ratios.

  • ndx – an Ndx object which define the list of trials to compute

  • feature_server – sidekit.FeaturesServer used to load the acoustic parameters

  • score_mat – a ndarray of scores to fill

  • seg_idx – the list of unique test segments to process. Those test segments should belong to the list of test segments in the ndx object. By setting seg_idx=None, all test segments from the ndx object will be processed