Updated model_iv.train() to use the sidekit.FactorAnalyser i-vector extractor

s4d/model_iv.py

@@ -20,7 +20,7 @@
 # You should have received a copy of the GNU Lesser General Public License
 # along with SIDEKIT.  If not, see <http://www.gnu.org/licenses/>.
 
-from sidekit import Mixture, StatServer, Scores, Ndx, PLDA_scoring, cosine_scoring, mahalanobis_scoring, two_covariance_scoring
+from sidekit import Mixture, StatServer, FactorAnalyser, Scores, Ndx, PLDA_scoring, cosine_scoring, mahalanobis_scoring, two_covariance_scoring
 from sidekit.sidekit_io import *
 import copy
 import numpy as np
@@ -67,12 +67,13 @@ class ModelIV:
             print('sn_cov: ', self.sn_cov.shape)
 
     def train(self, feature_server, idmap, normalization=True):
-        #stat = StatServer(idmap, self.ubm)
         stat = StatServer(idmap, distrib_nb=self.ubm.distrib_nb(), feature_size=self.ubm.dim()) 
         stat.accumulate_stat(ubm=self.ubm, feature_server=feature_server, seg_indices=range(stat.segset.shape[0]), num_thread=self.nb_thread)
-
         stat = stat.sum_stat_per_model()[0]
-        self.ivectors = stat.estimate_hidden(self.tv_mean, self.tv_sigma, V=self.tv, U=None, D=None, num_thread=self.nb_thread)[0]
+        
+        fa = FactorAnalyser(mean=self.tv_mean, Sigma=self.tv_sigma, F=self.tv)
+        self.ivectors = fa.extract_ivectors_single(self.ubm, stat)
+        
         if normalization:
             self.ivectors.spectral_norm_stat1(self.norm_mean[:1], self.norm_cov[:1])