0:00:16 | i or the u s and model |
---|
0:00:22 | a target o k |
---|
0:00:26 | the weighting rule |
---|
0:00:29 | a problem in the unimodal |
---|
0:00:35 | i all right |
---|
0:00:40 | okay |
---|
0:00:41 | i'm in the table |
---|
0:00:43 | a little or it all i mean how men |
---|
0:00:49 | and |
---|
0:00:51 | to improve the performance you know i wrote you might not |
---|
0:00:59 | it better |
---|
0:01:00 | but the is better |
---|
0:01:03 | we are currently are really hartman and recognition |
---|
0:01:08 | okay tomorrow i in rate in one or more |
---|
0:01:12 | i don't okay |
---|
0:01:14 | you map or |
---|
0:01:16 | a little more to the global mean not |
---|
0:01:21 | a more or the u i |
---|
0:01:27 | the problem we will have |
---|
0:01:31 | i |
---|
0:01:35 | we gotta identification |
---|
0:01:38 | and the |
---|
0:01:39 | a reasonable that the of course i |
---|
0:01:45 | although |
---|
0:01:46 | a new |
---|
0:01:48 | notation o a whole number one |
---|
0:01:51 | but i three right okay |
---|
0:01:55 | general model unit |
---|
0:01:58 | i write my |
---|
0:02:03 | the other hand |
---|
0:02:05 | or you might add more efficient model |
---|
0:02:11 | probably not you know there |
---|
0:02:19 | what the motivation for |
---|
0:02:22 | no i nor my usual the problem all the speaker recognition |
---|
0:02:29 | on the future |
---|
0:02:30 | only we use when i |
---|
0:02:34 | the are you speaker |
---|
0:02:38 | nothing better than |
---|
0:02:41 | a dimension |
---|
0:02:43 | well i not information in |
---|
0:02:46 | the core condition |
---|
0:02:50 | really |
---|
0:02:52 | well i |
---|
0:02:55 | the mobile |
---|
0:02:57 | okay well when i don't recognition model |
---|
0:03:04 | the model |
---|
0:03:05 | okay well |
---|
0:03:07 | i |
---|
0:03:08 | convolutional layer it more data |
---|
0:03:12 | the speaker model |
---|
0:03:15 | most people do you know well |
---|
0:03:18 | we more |
---|
0:03:21 | other than i'm recognition model |
---|
0:03:24 | a the fact |
---|
0:03:26 | independently |
---|
0:03:27 | and to get |
---|
0:03:31 | a little better than the one point five k u one and recognition |
---|
0:03:40 | i mean that no the one that |
---|
0:03:43 | it is not who and the in domain and with it |
---|
0:03:52 | well |
---|
0:03:53 | i |
---|
0:03:57 | in all the speaker and you know |
---|
0:04:00 | a whole and out in a |
---|
0:04:05 | if you allusion and what it is you |
---|
0:04:08 | well |
---|
0:04:13 | of the movie |
---|
0:04:14 | a more efficient addition there then you have the time |
---|
0:04:21 | in addition there than l |
---|
0:04:23 | well information open condition |
---|
0:04:28 | you know the |
---|
0:04:31 | the in the morning edition |
---|
0:04:34 | a model adaptation okay really global condition |
---|
0:04:38 | a eight conversation edition |
---|
0:04:42 | and the |
---|
0:04:44 | we have only okay |
---|
0:04:49 | on the other hand recognition |
---|
0:04:52 | i don't |
---|
0:04:53 | a really and i will be a really by a all out of domain |
---|
0:05:02 | well i and you the mac layer generating the n r |
---|
0:05:10 | the calibration they the i |
---|
0:05:13 | a binary okay or more |
---|
0:05:18 | the little i |
---|
0:05:21 | well by the channel estimate |
---|
0:05:26 | on the other habitation |
---|
0:05:29 | other application had available |
---|
0:05:34 | a be a |
---|
0:05:35 | but we directly are really apply channel |
---|
0:05:40 | a where we have we are counting on timit |
---|
0:05:47 | i between the core condition |
---|
0:05:52 | the idea is to a you know not to buy or by two |
---|
0:06:01 | i'm the weight |
---|
0:06:03 | i really |
---|
0:06:05 | using a not very i don't like to buy is |
---|
0:06:10 | a visual word by |
---|
0:06:13 | you the |
---|
0:06:14 | the application i |
---|
0:06:17 | i'm from the speaker who |
---|
0:06:20 | we |
---|
0:06:22 | at a |
---|
0:06:23 | okay the handle because we have i |
---|
0:06:33 | and |
---|
0:06:36 | probably okay |
---|
0:06:38 | the |
---|
0:06:42 | enrollment data about |
---|
0:06:46 | i o |
---|
0:06:49 | i u |
---|
0:06:51 | either the |
---|
0:06:52 | a longer than that |
---|
0:06:55 | where you that you are you |
---|
0:07:00 | e o to o e |
---|
0:07:03 | every meeting recognition |
---|
0:07:06 | although the without a |
---|
0:07:10 | and the |
---|
0:07:12 | i know |
---|
0:07:14 | i don't you do |
---|
0:07:17 | and the other hand but |
---|
0:07:21 | only one done by the okay |
---|
0:07:26 | okay in that |
---|
0:07:28 | well |
---|
0:07:29 | well everyone to you to the |
---|
0:07:33 | a e |
---|
0:07:35 | and okay you did not |
---|
0:07:38 | we have |
---|
0:07:39 | the or |
---|
0:07:47 | i |
---|
0:07:49 | but i |
---|
0:07:51 | the |
---|
0:07:51 | speaker identification |
---|
0:07:53 | you |
---|
0:07:55 | the one approach |
---|
0:07:59 | a that a whole training and that |
---|
0:08:04 | i don't know there is no longer than the other hand but |
---|
0:08:10 | a menu that and y two k |
---|
0:08:16 | okay and that |
---|
0:08:17 | okay |
---|
0:08:19 | it |
---|
0:08:20 | one okay |
---|
0:08:23 | you know to do the whole problem |
---|
0:08:27 | i don't know that will be used to a well known |
---|
0:08:32 | did you are at the not and not you |
---|
0:08:37 | are you that |
---|
0:08:40 | only option |
---|
0:08:42 | we you know |
---|
0:08:43 | i but the i and i'll |
---|
0:08:48 | i for the |
---|
0:08:49 | but the it that |
---|
0:08:53 | and you as a nice improvement |
---|
0:08:56 | well |
---|
0:08:57 | and i can be i |
---|
0:09:01 | and maybe a data |
---|
0:09:04 | a well you all the all pro |
---|
0:09:11 | and there is a okay i'm weight and the |
---|
0:09:16 | or you a bit |
---|
0:09:19 | and i'm time and then the more |
---|
0:09:23 | you have the right |
---|
0:09:25 | and the other p |
---|
0:09:31 | okay |
---|
0:09:34 | and one speaker recognition |
---|
0:09:37 | there are totally |
---|
0:09:39 | really and then |
---|
0:09:43 | okay |
---|
0:09:45 | a fusion of it or not |
---|
0:09:50 | well |
---|
0:09:52 | and then than a the thing but okay not only |
---|
0:09:58 | no value |
---|
0:10:00 | a very |
---|
0:10:02 | and that |
---|
0:10:04 | i just a the university human |
---|
0:10:09 | i'm not known |
---|
0:10:14 | well |
---|
0:10:23 | i |
---|
0:10:24 | that |
---|
0:10:25 | and it will be |
---|
0:10:29 | the i e |
---|
0:10:32 | we then |
---|
0:10:33 | i'm getting higher than i would be other |
---|
0:10:38 | and i three |
---|
0:10:39 | a speaker verification and you the speaker recognition model only |
---|
0:10:45 | i e |
---|
0:10:47 | i e |
---|
0:10:48 | are there |
---|
0:10:51 | data a little |
---|
0:10:53 | and the speaker recognition more |
---|
0:10:55 | a using them |
---|
0:11:01 | i right |
---|
0:11:02 | i e |
---|
0:11:04 | i four or more using a total and one iteration |
---|
0:11:09 | well you |
---|
0:11:11 | and i e |
---|
0:11:13 | okay we have at |
---|
0:11:15 | the i a model that using a total between and among the more you |
---|
0:11:22 | the recognition |
---|
0:11:25 | i'm the of the more like a you know how you know what |
---|
0:11:31 | i am with organic probably layer |
---|
0:11:35 | we don't well |
---|
0:11:37 | i mean all |
---|
0:11:39 | a little the |
---|
0:11:41 | well |
---|
0:11:44 | and a |
---|
0:11:45 | okay the problem of the one level |
---|
0:11:50 | i |
---|
0:11:51 | it might have an error rate |
---|
0:11:55 | a in well i don't really condition |
---|
0:11:59 | and order to make the model i to make a movie that it would be |
---|
0:12:06 | and the television if you |
---|
0:12:09 | a it's a convolutional |
---|
0:12:12 | well |
---|
0:12:14 | okay |
---|
0:12:15 | and then combine them here |
---|
0:12:17 | there are more efficient |
---|
0:12:21 | and the actual the model |
---|
0:12:27 | and do not well what is that it may be a problem |
---|
0:12:35 | well |
---|
0:12:37 | i the them you |
---|
0:12:39 | the we didn't use a suitable |
---|
0:12:43 | but you wanna you know what |
---|
0:12:47 | and that we the problem you |
---|
0:12:54 | and the |
---|
0:12:55 | but there is no |
---|
0:12:57 | the |
---|
0:13:00 | it'll |
---|
0:13:01 | well |
---|
0:13:05 | you the more you rate |
---|
0:13:11 | i e |
---|
0:13:13 | i performance of the extended edition |
---|
0:13:17 | a |
---|
0:13:19 | the united |
---|
0:13:25 | and |
---|
0:13:28 | i |
---|
0:13:28 | i and i |
---|
0:13:30 | and i e and i that the |
---|
0:13:33 | and you want to than what we implement a one |
---|
0:13:42 | a very good that you only not |
---|
0:13:47 | how to i i e |
---|
0:13:50 | and you know data model |
---|
0:13:53 | she'll |
---|
0:13:54 | by a nice improvement |
---|
0:13:56 | you know a little you that'll |
---|
0:14:00 | and i probability a really all you can |
---|
0:14:06 | i |
---|
0:14:07 | the information and okay |
---|
0:14:13 | i |
---|
0:14:14 | you know well |
---|
0:14:18 | and then |
---|
0:14:19 | but the |
---|
0:14:22 | recognition |
---|
0:14:23 | a lot of the |
---|
0:14:27 | no |
---|
0:14:28 | the |
---|
0:14:29 | here |
---|
0:14:31 | but not or you |
---|
0:14:35 | from people i and so that |
---|
0:14:39 | through a piranha |
---|
0:14:41 | the individual model |
---|
0:14:47 | if i the you know little or you can perform i might not |
---|
0:14:55 | right |
---|
0:14:56 | you can we try to model i s o a long |
---|
0:15:02 | having to do not |
---|
0:15:05 | them |
---|
0:15:06 | relative improvement okay |
---|
0:15:08 | in r e i e |
---|
0:15:11 | i e at a |
---|
0:15:16 | but the more |
---|
0:15:21 | for the whole speaker recognition |
---|
0:15:25 | a little i know the true value then problem the model |
---|
0:15:33 | it completely okay on a |
---|
0:15:37 | v you in a further more |
---|
0:15:47 | there you go fishing model you years you model and a like i of it |
---|
0:15:55 | you more normal or |
---|
0:15:59 | then there is a little more i |
---|
0:16:05 | the i |
---|
0:16:07 | only the that |
---|
0:16:09 | the biphone like |
---|
0:16:11 | i don't you |
---|
0:16:12 | okay i |
---|
0:16:14 | you are |
---|
0:16:16 | right |
---|
0:16:22 | you tomorrow you are model |
---|
0:16:25 | my anymore or and then |
---|
0:16:29 | the women |
---|
0:16:30 | information |
---|
0:16:31 | and you have two more years |
---|
0:16:35 | what do you all |
---|
0:16:42 | and the three the original the real identification |
---|
0:16:49 | and the |
---|
0:16:51 | i i'm at |
---|
0:16:53 | but i thirty and i |
---|
0:17:00 | in a file |
---|
0:17:03 | you know the a i e i'm sorry |
---|
0:17:08 | i don't know that or a i u i e |
---|
0:17:17 | of the plan |
---|
0:17:21 | well |
---|
0:17:23 | a mission it you know that only you the a i e and but i |
---|
0:17:29 | e |
---|
0:17:30 | and the barrel |
---|
0:17:32 | the new combination the in narrative on the that |
---|
0:17:37 | at the end and i |
---|
0:17:40 | it is clear from the |
---|
0:17:42 | that are not |
---|
0:17:45 | a final no |
---|
0:17:47 | bigger |
---|
0:17:48 | i you know an additional |
---|
0:17:51 | the video i e |
---|
0:17:55 | if you're gonna i and |
---|
0:17:58 | and i and i |
---|
0:18:00 | a the i know |
---|
0:18:04 | well you |
---|
0:18:05 | i |
---|
0:18:06 | you |
---|
0:18:08 | i |
---|
0:18:09 | the |
---|
0:18:11 | recognition |
---|
0:18:12 | although |
---|
0:18:14 | on the number three |
---|
0:18:18 | a the are really article |
---|
0:18:26 | the level |
---|
0:18:28 | and the by i i'm the colour |
---|
0:18:36 | the p and |
---|
0:18:43 | i didn't |
---|
0:18:45 | on the paper |
---|
0:18:47 | a lot of data by the |
---|
0:18:50 | six hub four |
---|
0:18:53 | and the unit for |
---|
0:18:56 | well |
---|
0:18:58 | you know the two |
---|
0:19:00 | cool |
---|
0:19:00 | the |
---|
0:19:02 | identification |
---|
0:19:03 | and that a simple |
---|
0:19:05 | the or not |
---|
0:19:10 | a possible |
---|
0:19:12 | there is more |
---|
0:19:15 | well you never the |
---|
0:19:19 | all the model |
---|
0:19:21 | and i |
---|
0:19:24 | the relevant information |
---|
0:19:27 | but the problem then you're able to spend the only you the graph model |
---|
0:19:34 | and the over |
---|
0:19:36 | okay |
---|
0:19:38 | i four u o a i-vector model kind |
---|
0:19:41 | what the |
---|
0:19:42 | you great |
---|
0:19:48 | or |
---|
0:19:50 | and future work |
---|
0:19:52 | that is that you the multi condition |
---|
0:19:55 | i can be really you |
---|
0:19:59 | and the what we're very well but |
---|
0:20:04 | a movie that i |
---|
0:20:06 | okay i four u |
---|
0:20:10 | you |
---|