0:00:16 | and i don't have circle |
---|---|

0:00:24 | i four u |

0:00:26 | the speaker recognition workshop |

0:00:32 | the decision or not |

0:00:37 | the right |

0:00:40 | each one |

0:00:43 | german |

0:00:45 | furthermore |

0:00:47 | e |

0:00:54 | i think |

0:00:56 | okay what the |

0:01:01 | or the u i i think and i think |

0:01:09 | right |

0:01:19 | in the u |

0:01:22 | on a greater and would be efficiently |

0:01:27 | can he may be paid on time nine cepstral coefficients |

0:01:34 | then you are using it can also report that you can be one or two |

0:01:49 | and |

0:01:51 | evaluation and you know |

0:01:55 | and then just |

0:02:00 | all right |

0:02:02 | i wouldn't can control only |

0:02:05 | i don't |

0:02:07 | and two |

0:02:10 | because you |

0:02:12 | section |

0:02:13 | finally and b |

0:02:21 | that is not required to read |

0:02:26 | right behind the original speaker |

0:02:30 | i to me |

0:02:34 | and |

0:02:36 | i |

0:02:38 | or not |

0:02:43 | i |

0:02:48 | however |

0:02:49 | it just one and evaluated on the next |

0:02:55 | this means that |

0:03:10 | right |

0:03:14 | and |

0:03:15 | phonotactic features |

0:03:19 | all right |

0:03:24 | why only r e one |

0:03:31 | it can be or finally we got to each of your equally well in that |

0:03:44 | one |

0:03:46 | in |

0:03:47 | in the gmm be in the u i |

0:03:52 | vectors |

0:03:55 | i the only a very |

0:03:59 | where and how i-vectors |

0:04:04 | i mean |

0:04:08 | on |

0:04:14 | it is |

0:04:15 | lda and you know |

0:04:27 | to provide a little time |

0:04:33 | that can be okay well or better |

0:04:41 | the main motivation behind |

0:04:49 | in one |

0:04:52 | i really |

0:04:56 | why the into it i think i might not be able to the we were |

0:05:10 | able |

0:05:11 | section four |

0:05:14 | there are derived from a wide open set of the well |

0:05:21 | in |

0:05:22 | i |

0:05:24 | in order |

0:05:32 | and |

0:05:38 | then |

0:05:41 | but development |

0:05:48 | finally you know |

0:05:53 | no |

0:05:53 | we can be different |

0:05:59 | based in it |

0:06:06 | the window |

0:06:09 | in an image |

0:06:12 | i and i |

0:06:18 | furthermore i |

0:06:20 | on the one |

0:06:22 | but |

0:06:26 | then the invention limited |

0:06:34 | i mean vector that f and g |

0:06:47 | a vector right in one |

0:06:52 | or what do not all the unit vector |

0:06:56 | well fourteen |

0:06:59 | and |

0:07:00 | i |

0:07:02 | one |

0:07:04 | sure that |

0:07:09 | i computation |

0:07:15 | the low |

0:07:18 | it |

0:07:21 | i |

0:07:23 | and |

0:07:26 | and i in |

0:07:30 | the problem |

0:07:32 | the |

0:07:33 | i mean |

0:07:34 | and all |

0:07:40 | i |

0:07:44 | and |

0:07:45 | wait |

0:07:47 | like image over time |

0:07:50 | finally the update |

0:07:54 | you're right |

0:07:55 | you date |

0:07:59 | okay |

0:08:05 | but i mean not |

0:08:11 | you anyone e |

0:08:15 | and one two |

0:08:17 | don't have been working |

0:08:19 | i don't want to |

0:08:24 | and |

0:08:26 | and only one working |

0:08:28 | computed |

0:08:31 | one |

0:08:34 | okay and option |

0:08:41 | one can be |

0:08:45 | and more original have to be a and you don't dimensional |

0:08:54 | i o compute |

0:08:59 | you are |

0:09:01 | i |

0:09:02 | and |

0:09:04 | for the computation |

0:09:12 | and that |

0:09:14 | it also |

0:09:20 | okay we have |

0:09:24 | one computation |

0:09:30 | i |

0:09:32 | do not need a one and e |

0:09:39 | i |

0:09:42 | i think |

0:09:46 | i mean and unit one |

0:09:50 | i |

0:09:51 | i |

0:09:53 | and a |

0:09:57 | one hundred |

0:09:59 | i |

0:10:00 | i-vectors |

0:10:03 | i |

0:10:10 | well no more than one might be |

0:10:15 | and one |

0:10:24 | what |

0:10:31 | and german |

0:10:34 | how well thank you |

0:10:39 | okay |

0:10:43 | and i |

0:10:44 | you and all three in |

0:10:50 | management for evaluation |

0:10:53 | e |

0:10:56 | we want you |

0:10:59 | i |

0:11:01 | and the way i |

0:11:03 | from i |

0:11:05 | sure |

0:11:09 | the i-vector |

0:11:12 | and then |

0:11:14 | and |

0:11:18 | for every |

0:11:19 | we can do |

0:11:21 | but i |

0:11:23 | two |

0:11:25 | eight |

0:11:27 | or not |

0:11:29 | and |

0:11:31 | in section |

0:11:33 | therefore important |

0:11:37 | for that |

0:11:39 | you or i |

0:11:43 | okay |

0:11:46 | i and i one |

0:11:51 | and in fact it can work on |

0:12:00 | in view |

0:12:03 | one o d baseline |

0:12:07 | and |

0:12:10 | what do you |

0:12:14 | i one |

0:12:19 | i dunno |

0:12:21 | or you know how one |

0:12:25 | and |

0:12:27 | women |

0:12:30 | and it can |

0:12:34 | the quantization |

0:12:37 | where are in trouble |

0:12:43 | i'm sure that you can be |

0:12:48 | one |

0:12:50 | the p |

0:12:51 | i |

0:12:56 | a company |

0:12:58 | i |

0:13:00 | okay and |

0:13:02 | on the gain |

0:13:03 | and |

0:13:04 | and england for to be able |

0:13:12 | and that we propose |

0:13:15 | you know |

0:13:19 | i four u |

0:13:23 | will be data |

0:13:25 | a big |

0:13:26 | the point two |

0:13:34 | in our |

0:13:40 | okay |

0:13:42 | in the evaluation data |

0:13:46 | based on the incline |

0:13:50 | after that you by the |

0:13:56 | i don't the problem |

0:13:59 | cool |

0:14:02 | all the time domain |

0:14:04 | in |

0:14:06 | rule |

0:14:07 | right |

0:14:13 | well i |

0:14:24 | i don't know that in detail |

0:14:27 | a woman and the right |

0:14:41 | what i don't |

0:14:43 | the whole |

0:14:46 | well |

0:14:48 | in room b |

0:14:55 | i don't |

0:15:00 | it didn't |

0:15:05 | from the number and can be |

0:15:08 | the problem |

0:15:13 | we can write a |

0:15:17 | that is |

0:15:18 | our balloon of entropy |

0:15:24 | my little or no you're enjoying of the can go into a low dimensional |

0:15:37 | right |

0:15:38 | in particular or |

0:15:51 | by really global warming well |

0:15:55 | i don't think that are more than |

0:16:05 | and so on |

0:16:10 | but i that time |

0:16:19 | and they are |

0:16:22 | our one |

0:16:25 | or can be a problem i-vector i four point one high dimensional |

0:16:46 | for all of them |

0:16:51 | and the i-vector |

0:16:54 | the do not |

0:16:58 | and i can be the most |

0:17:02 | i don't |

0:17:04 | of course |

0:17:10 | for example |

0:17:12 | in the back one |

0:17:17 | quite good |

0:17:18 | equation one point one |

0:17:27 | and hundred and an i-vector |

0:17:29 | the only one hundred and d o b i |

0:17:43 | that was used |

0:17:47 | have you can be done |

0:17:53 | okay |

0:17:54 | and i mean and i think i four |

0:18:00 | and you |

0:18:01 | well |

0:18:03 | that you can be done on a and b |

0:18:14 | by |

0:18:15 | one working |

0:18:17 | i four point four in can perform better than four i-vectors |

0:18:32 | and i can |

0:18:34 | i |

0:18:37 | i that |

0:18:42 | and what have one |

0:18:47 | alright |

0:18:49 | one |

0:18:56 | when they i and i know how much of the wood |

0:19:08 | and one time |

0:19:15 | i one and |

0:19:19 | where |

0:19:21 | in performance |

0:19:28 | i |

0:19:29 | very well |

0:19:31 | and that kind of a lot of iraqi |

0:19:38 | and you |