0:00:15although an afternoon
0:00:18it's a problem i mean every to mister technology later to present my paper this
0:00:26so what the problem is a sampling to address the problem of variability in the
0:00:31i-vector space you to the acoustic content of the
0:00:37and the language is the main one of the main now
0:00:40source of this variability
0:00:43the probabilistic linear discriminant analysis while going to model the kind of source variability but
0:00:51it cannot some model this variability using multilingual without a multilingual apartments that's for each
0:01:03there is a one a method called language normalized w c n
0:01:08which is designed to model this variability by extending the source normalized mfcc
0:01:17this is done before a prior to the p lda training camp
0:01:21so what i am going to post is to
0:01:26propose a purely training algorithm
0:01:29we would be built to reduce this language in fact
0:01:33so by estimating the speaker and channel subspace stuff from multilingual utterances your we can
0:01:40be appealed it can be able to work independent
0:01:47so when evaluated on the nist sre two thousand eight core condition
0:01:53we were able to you know
0:01:57reduce the fact
0:01:59we use the russian spanish arabic and mandarin in addition to english
0:02:07so in comparison with the baseline system we use double system was we were able
0:02:14we choose the language effects by ten percent the equal error rates
0:02:22so that's it