| 0:00:14 | have a we all my name on the know |
|---|
| 0:00:18 | i do right to well the of the initial model control whole |
|---|
| 0:00:21 | generalized |
|---|
| 0:00:22 | sidelobe canceller based |
|---|
| 0:00:24 | speech in is not |
|---|
| 0:00:30 | i i'm going to spray might talking to |
|---|
| 0:00:32 | for parts |
|---|
| 0:00:34 | or |
|---|
| 0:00:38 | oh i one wanna |
|---|
| 0:00:39 | well starts by explaining the |
|---|
| 0:00:43 | is spring that you S C and |
|---|
| 0:00:46 | at a M I and |
|---|
| 0:00:48 | oh it's do that i use that is |
|---|
| 0:00:50 | the proposed amc |
|---|
| 0:00:52 | for each |
|---|
| 0:00:53 | i |
|---|
| 0:00:54 | the duration of arrival or based the or |
|---|
| 0:00:58 | but to to a lower of channel mean type type to montana hundred much a |
|---|
| 0:01:03 | estimation um |
|---|
| 0:01:04 | oh is explored |
|---|
| 0:01:06 | and |
|---|
| 0:01:08 | the a target directional speech of cells of progress |
|---|
| 0:01:11 | probability estimation used also explored |
|---|
| 0:01:14 | and |
|---|
| 0:01:16 | the pipe parents |
|---|
| 0:01:17 | oh |
|---|
| 0:01:18 | the proposed algorithm we use the at mean in and i use summarised to my study |
|---|
| 0:01:27 | or |
|---|
| 0:01:28 | are sent to a much channel based the speech enhancement |
|---|
| 0:01:32 | oh approach is a a cut to as three into these i i've pass |
|---|
| 0:01:37 | oh pimp or yeah well |
|---|
| 0:01:39 | yes this |
|---|
| 0:01:40 | and masking approach you and class are and but china |
|---|
| 0:01:43 | the the pretend |
|---|
| 0:01:45 | i |
|---|
| 0:01:47 | or or other than oh |
|---|
| 0:01:50 | in he's he's a nineteen is a pretty issue need up over there |
|---|
| 0:01:55 | oh |
|---|
| 0:01:56 | by P |
|---|
| 0:01:57 | that's to or approaching proteins |
|---|
| 0:01:59 | a plus the in |
|---|
| 0:02:01 | this part |
|---|
| 0:02:04 | so |
|---|
| 0:02:05 | so this is she or |
|---|
| 0:02:07 | in |
|---|
| 0:02:08 | compares the two or picks the problem |
|---|
| 0:02:12 | and other the meant to use and those can signal |
|---|
| 0:02:15 | oh at at bit proof of peaks from calm uh a a at the bit |
|---|
| 0:02:20 | or for a as |
|---|
| 0:02:22 | that takes that in |
|---|
| 0:02:23 | in the type |
|---|
| 0:02:25 | to options so that |
|---|
| 0:02:26 | no that or so as you can also so to obtain a at to and |
|---|
| 0:02:31 | in contrast |
|---|
| 0:02:33 | let look and products |
|---|
| 0:02:35 | but i don't |
|---|
| 0:02:36 | to noise show can us so that |
|---|
| 0:02:39 | whole |
|---|
| 0:02:39 | not have |
|---|
| 0:02:40 | to a chance of the nose |
|---|
| 0:02:42 | mean that a collection can thus |
|---|
| 0:02:45 | oh can pass through |
|---|
| 0:02:46 | and |
|---|
| 0:02:48 | my a cancer that generates |
|---|
| 0:02:50 | but you know as good at the processing and that's it's probably a and it be a map of signals |
|---|
| 0:02:56 | so |
|---|
| 0:02:58 | yeah |
|---|
| 0:02:58 | use uh |
|---|
| 0:02:59 | oh you do one |
|---|
| 0:03:01 | and and don't was set |
|---|
| 0:03:03 | but the lead to by reason is |
|---|
| 0:03:06 | for a two |
|---|
| 0:03:06 | in the uh yeah and and C |
|---|
| 0:03:09 | or |
|---|
| 0:03:13 | the main problem yeah but to a solution you is |
|---|
| 0:03:16 | or |
|---|
| 0:03:18 | how do it |
|---|
| 0:03:18 | to a show together or |
|---|
| 0:03:21 | well |
|---|
| 0:03:23 | oh have a to assess about special |
|---|
| 0:03:26 | um |
|---|
| 0:03:29 | especially uh |
|---|
| 0:03:31 | but |
|---|
| 0:03:33 | and they should have had attention |
|---|
| 0:03:35 | mission as you do a petition that's gonna come from that scene everybody a |
|---|
| 0:03:39 | oh to give us performance so |
|---|
| 0:03:41 | use |
|---|
| 0:03:42 | a just a C P |
|---|
| 0:03:43 | is that there is just a touch and others |
|---|
| 0:03:46 | come to |
|---|
| 0:03:47 | the |
|---|
| 0:03:48 | get that processing that it would be have and the S C |
|---|
| 0:03:52 | i i have to the do it's again a to been to both |
|---|
| 0:03:57 | so that was assume is cut can this to time |
|---|
| 0:04:01 | uh |
|---|
| 0:04:02 | hi but this stone and so that the is a |
|---|
| 0:04:05 | uh how the addition is |
|---|
| 0:04:07 | there's is um |
|---|
| 0:04:08 | oh have that |
|---|
| 0:04:09 | how that sparse component is for me not be it's clear to be in a not so |
|---|
| 0:04:14 | how this is for about a |
|---|
| 0:04:16 | the |
|---|
| 0:04:18 | yeah yeah |
|---|
| 0:04:20 | sure is |
|---|
| 0:04:22 | so that have a nice |
|---|
| 0:04:23 | how a decision based M C |
|---|
| 0:04:26 | a |
|---|
| 0:04:27 | and to that and is the based them to do and |
|---|
| 0:04:30 | where the a should be can target |
|---|
| 0:04:33 | and there's a clear |
|---|
| 0:04:34 | in from us |
|---|
| 0:04:36 | yeah a separate this is based on the |
|---|
| 0:04:40 | how much the time the speech energy is included in the in |
|---|
| 0:04:43 | to have yeah |
|---|
| 0:04:44 | a time-frequency bin |
|---|
| 0:04:46 | so |
|---|
| 0:04:47 | it's separate addition for this |
|---|
| 0:04:49 | the |
|---|
| 0:04:49 | in the where the tears yeah it's important to to run |
|---|
| 0:04:54 | a a no some put additional approach |
|---|
| 0:04:56 | is user |
|---|
| 0:04:58 | it to as |
|---|
| 0:04:58 | that kind of |
|---|
| 0:05:00 | speech |
|---|
| 0:05:01 | that's that's is as probabilities |
|---|
| 0:05:05 | so this so that a |
|---|
| 0:05:06 | this so that the shows |
|---|
| 0:05:08 | or or i or proposed to sort additionally am |
|---|
| 0:05:11 | oh and i to the conventional or and a she or |
|---|
| 0:05:15 | in i |
|---|
| 0:05:16 | a the area she |
|---|
| 0:05:19 | or of a probability a |
|---|
| 0:05:21 | but |
|---|
| 0:05:22 | the clutter probability in |
|---|
| 0:05:25 | improved the you a cancer uh and |
|---|
| 0:05:28 | i have a the probability is improve that you know |
|---|
| 0:05:31 | i the can the kinetics |
|---|
| 0:05:33 | uh |
|---|
| 0:05:34 | they have a that |
|---|
| 0:05:35 | for a a point of high and of the probability use the plan to ask a product of |
|---|
| 0:05:40 | or require a and clear |
|---|
| 0:05:42 | but known at least |
|---|
| 0:05:43 | in our study |
|---|
| 0:05:48 | so |
|---|
| 0:05:54 | so missed stats |
|---|
| 0:05:55 | or |
|---|
| 0:05:57 | you start from to help the is |
|---|
| 0:05:59 | or |
|---|
| 0:06:00 | that this thing is |
|---|
| 0:06:02 | no permutation of components |
|---|
| 0:06:04 | and |
|---|
| 0:06:05 | this and is kind of the components so |
|---|
| 0:06:09 | yeah nice to have but is time |
|---|
| 0:06:11 | to have to a |
|---|
| 0:06:13 | you can try |
|---|
| 0:06:15 | the |
|---|
| 0:06:16 | i do to action a speech to has i could mean have to that |
|---|
| 0:06:20 | and |
|---|
| 0:06:21 | for um |
|---|
| 0:06:22 | that |
|---|
| 0:06:23 | oh like or you the ratio that we can obtain the core in a probability so i could be |
|---|
| 0:06:33 | a |
|---|
| 0:06:34 | meaning to i don't tuition are |
|---|
| 0:06:37 | components |
|---|
| 0:06:39 | at that we we need to estimate the perry so the kind of all components in target directional components so |
|---|
| 0:06:46 | the estimation process to use of i could do this |
|---|
| 0:06:49 | the you venture |
|---|
| 0:06:51 | uh a can have that the |
|---|
| 0:06:54 | i |
|---|
| 0:06:54 | uh |
|---|
| 0:06:56 | a from a uh had a a a for a to have a use that to probably for the reliable |
|---|
| 0:07:01 | estimate of to |
|---|
| 0:07:03 | and it to action a speech of or pretty probabilities |
|---|
| 0:07:12 | uh |
|---|
| 0:07:14 | this parameter one |
|---|
| 0:07:16 | look at this power but that uh |
|---|
| 0:07:21 | use the pin show is have that |
|---|
| 0:07:23 | direction or |
|---|
| 0:07:24 | to not how to a directional components |
|---|
| 0:07:27 | uh |
|---|
| 0:07:28 | no |
|---|
| 0:07:29 | do a tune and tied it to show that component |
|---|
| 0:07:32 | issue |
|---|
| 0:07:33 | oh |
|---|
| 0:07:34 | is is use the can show you that's and clear that or a word special a |
|---|
| 0:07:39 | information |
|---|
| 0:07:40 | so |
|---|
| 0:07:41 | hmmm |
|---|
| 0:07:42 | in prayer |
|---|
| 0:07:43 | you improved to |
|---|
| 0:07:44 | and an additional uh a no |
|---|
| 0:07:46 | by incorporating a |
|---|
| 0:07:48 | but you it is gonna press |
|---|
| 0:07:50 | to |
|---|
| 0:07:51 | T that estimate |
|---|
| 0:07:53 | oh i could do this is up where this is opposed to the scope was to plano |
|---|
| 0:08:00 | the million |
|---|
| 0:08:01 | you |
|---|
| 0:08:02 | to |
|---|
| 0:08:03 | estimate the the of us to have an estimate |
|---|
| 0:08:12 | so that stats |
|---|
| 0:08:14 | most type |
|---|
| 0:08:15 | a problem |
|---|
| 0:08:16 | the to been can extract to this that to go |
|---|
| 0:08:20 | but channel in from up to |
|---|
| 0:08:24 | so this is that in problems |
|---|
| 0:08:26 | storage to a problem |
|---|
| 0:08:29 | a this that so we can |
|---|
| 0:08:31 | i right |
|---|
| 0:08:32 | that pimp and the estimates condition |
|---|
| 0:08:35 | oh |
|---|
| 0:08:36 | this to me yes |
|---|
| 0:08:37 | but it it about the function to |
|---|
| 0:08:39 | improvements the |
|---|
| 0:08:41 | the the data base |
|---|
| 0:08:44 | you see creation use |
|---|
| 0:08:46 | uh |
|---|
| 0:08:47 | can be you to this |
|---|
| 0:08:49 | or |
|---|
| 0:08:51 | at this occasion |
|---|
| 0:08:52 | so we can try to the |
|---|
| 0:08:54 | oh |
|---|
| 0:08:55 | you can do that to knots compunction to be formers |
|---|
| 0:08:58 | right could these |
|---|
| 0:09:00 | do you do this you can and me |
|---|
| 0:09:03 | can can it |
|---|
| 0:09:04 | had to H are not you know a so and there's a comparison like this |
|---|
| 0:09:08 | uh of that kind of direction and that's fashion |
|---|
| 0:09:11 | and me yes |
|---|
| 0:09:12 | oh to the for mapping |
|---|
| 0:09:14 | and can promote |
|---|
| 0:09:15 | and for the image compression |
|---|
| 0:09:18 | we you use spread phonetics |
|---|
| 0:09:20 | so |
|---|
| 0:09:22 | me |
|---|
| 0:09:23 | obtain a the hundred to actually |
|---|
| 0:09:26 | sure |
|---|
| 0:09:27 | that's component yeah |
|---|
| 0:09:28 | and i don't need to completion |
|---|
| 0:09:30 | or like this |
|---|
| 0:09:35 | so you know to that i to two action and is the can question we |
|---|
| 0:09:39 | can you |
|---|
| 0:09:41 | we can obtain the approximate but that approximately the but don't know the kind of direction ah |
|---|
| 0:09:48 | and of to the now and then by the collection of components like but these |
|---|
| 0:09:52 | the |
|---|
| 0:09:53 | that's can is applied to a |
|---|
| 0:09:56 | yeah wrapper as channel signal signals |
|---|
| 0:09:58 | so you doing this |
|---|
| 0:10:01 | these approximately do |
|---|
| 0:10:03 | which signals maybe |
|---|
| 0:10:05 | obtained |
|---|
| 0:10:06 | with the |
|---|
| 0:10:07 | a close to can a one channels |
|---|
| 0:10:09 | i could be a |
|---|
| 0:10:12 | the |
|---|
| 0:10:13 | the corpus to the can only |
|---|
| 0:10:16 | uh |
|---|
| 0:10:18 | but a it's somewhere else to the just to |
|---|
| 0:10:21 | to to has a function like these |
|---|
| 0:10:24 | so so if i in it can obtain |
|---|
| 0:10:27 | or |
|---|
| 0:10:28 | a post so it an i by averaging the value |
|---|
| 0:10:37 | a |
|---|
| 0:10:38 | up to a poor |
|---|
| 0:10:39 | um |
|---|
| 0:10:40 | a to a like open or E |
|---|
| 0:10:43 | C send to per cent you may take to balance space |
|---|
| 0:10:46 | use is constructed |
|---|
| 0:10:48 | you know a you might amounts |
|---|
| 0:10:50 | and speech and energy G uh from we could do |
|---|
| 0:10:54 | and |
|---|
| 0:10:55 | actually mixed need of |
|---|
| 0:10:56 | the purpose in on it |
|---|
| 0:11:03 | a you prepared the |
|---|
| 0:11:05 | to additional provisions for the performance comparison of you |
|---|
| 0:11:09 | no you reduction in every and i and the of china evaluation of speech quality or |
|---|
| 0:11:14 | for the had a dish ten |
|---|
| 0:11:16 | that of |
|---|
| 0:11:17 | to as you can interval was may now to detect key |
|---|
| 0:11:20 | and |
|---|
| 0:11:21 | for about a conventional a separate the decisions are we |
|---|
| 0:11:25 | for a |
|---|
| 0:11:27 | i pitch |
|---|
| 0:11:28 | you know |
|---|
| 0:11:29 | a in the |
|---|
| 0:11:31 | let ones |
|---|
| 0:11:32 | paper |
|---|
| 0:11:34 | yeah |
|---|
| 0:11:35 | uh proposed the |
|---|
| 0:11:36 | what this is a message |
|---|
| 0:11:43 | so |
|---|
| 0:11:48 | C so is a noisy signal |
|---|
| 0:11:50 | yeah |
|---|
| 0:11:53 | it is quite a and it means all the white house |
|---|
| 0:11:57 | is |
|---|
| 0:11:58 | okay |
|---|
| 0:11:58 | but |
|---|
| 0:12:02 | and |
|---|
| 0:12:03 | uh uh of the decision |
|---|
| 0:12:05 | but |
|---|
| 0:12:07 | a |
|---|
| 0:12:09 | yeah |
|---|
| 0:12:11 | proposed |
|---|
| 0:12:11 | so we show |
|---|
| 0:12:13 | but |
|---|
| 0:12:16 | and and in the menu or lead the T |
|---|
| 0:12:19 | is that have the least still |
|---|
| 0:12:23 | it just the that runs |
|---|
| 0:12:25 | yeah |
|---|
| 0:12:32 | so |
|---|
| 0:12:33 | no you reduction and propose |
|---|
| 0:12:35 | evaluation |
|---|
| 0:12:37 | oh about |
|---|
| 0:12:38 | the uh and that the noise reduction was used to compare just signal of that just not just gonna of |
|---|
| 0:12:43 | new the than a discussion |
|---|
| 0:12:46 | the |
|---|
| 0:12:47 | in the |
|---|
| 0:12:51 | yes |
|---|
| 0:12:51 | or this proposed the |
|---|
| 0:12:53 | or |
|---|
| 0:12:54 | approach |
|---|
| 0:12:55 | or |
|---|
| 0:12:56 | have a that's sure |
|---|
| 0:12:57 | and it has |
|---|
| 0:12:59 | that the performance |
|---|
| 0:13:00 | that are our results |
|---|
| 0:13:05 | or was so in that every test in uh |
|---|
| 0:13:07 | or pulse |
|---|
| 0:13:11 | or our proposed approach |
|---|
| 0:13:13 | uh compared the performance |
|---|
| 0:13:15 | to |
|---|
| 0:13:17 | then you only had a distant and |
|---|
| 0:13:19 | and the parents then |
|---|
| 0:13:20 | conduct of the addition |
|---|
| 0:13:26 | and |
|---|
| 0:13:27 | this an issue |
|---|
| 0:13:29 | the pesq can has |
|---|
| 0:13:31 | yeah |
|---|
| 0:13:34 | the proposed that amc a |
|---|
| 0:13:35 | oh |
|---|
| 0:13:36 | has |
|---|
| 0:13:37 | or better performance |
|---|
| 0:13:39 | or compare with people most than the other |
|---|
| 0:13:43 | oh yeah |
|---|
| 0:13:44 | commissioner soap addition am C |
|---|
| 0:13:51 | so in conclusion in conclusion |
|---|
| 0:13:53 | a a plus to a to do this the post to train a was |
|---|
| 0:13:58 | in this can improve the reliability here |
|---|
| 0:14:01 | oh a lot of at time |
|---|
| 0:14:03 | to the city P S |
|---|
| 0:14:05 | and |
|---|
| 0:14:06 | however i have a the problem |
|---|
| 0:14:09 | a decision is was to |
|---|
| 0:14:11 | yeah i in she maybe yeah |
|---|
| 0:14:13 | and can of up to this show |
|---|
| 0:14:16 | okay i that further the to use in providing that and see no it has a consumer |
|---|
| 0:14:21 | so |
|---|
| 0:14:22 | it that's shown per the performance |
|---|
| 0:14:24 | compares them using pesq a and a every scenario |
|---|
| 0:14:29 | uh uh uh uh that |
|---|
| 0:14:31 | i for for the was this time to |
|---|
| 0:14:33 | but from better than |
|---|
| 0:14:35 | that |
|---|
| 0:14:35 | conventional |
|---|
| 0:14:37 | a a a a a to see this |
|---|
| 0:14:42 | you |
|---|
| 0:14:43 | thank you very much |
|---|
| 0:14:47 | i and a questions comments |
|---|
| 0:14:50 | from the audience |
|---|
| 0:14:52 | yes please |
|---|
| 0:14:54 | a a is a microphone here |
|---|
| 0:15:03 | i |
|---|
| 0:15:03 | yeah i wanted to ask "'em" your own experiments in the car |
|---|
| 0:15:07 | it was it did some quite good but you played that the and what i would be interested if you |
|---|
| 0:15:11 | make different experiments with those will then windows can would |
|---|
| 0:15:15 | in on and you know file be on one P um and things like that i mean |
|---|
| 0:15:20 | but that but it but that you try to do that at the critical points of work |
|---|
| 0:15:24 | oh of of your to the menu been dealing with the in situation |
|---|
| 0:15:28 | a a a a uh i just the can the noise |
|---|
| 0:15:31 | the |
|---|
| 0:15:32 | maybe stay or noise |
|---|
| 0:15:35 | problem |
|---|
| 0:15:36 | "'cause" to system |
|---|
| 0:15:38 | so does it mean that the engine was not one to two |
|---|
| 0:15:40 | yes i think and |
|---|
| 0:15:42 | that kind of sees |
|---|
| 0:15:46 | oh |
|---|
| 0:15:47 | is |
|---|
| 0:15:49 | i you the from that of conventional we know |
|---|
| 0:15:51 | two |
|---|
| 0:15:52 | i see one |
|---|
| 0:15:54 | and you |
|---|
| 0:15:56 | so and how many microphones yeah yeah for this application |
|---|
| 0:15:59 | well |
|---|
| 0:16:00 | and and microphones that we have there |
|---|
| 0:16:02 | how many might microphones |
|---|
| 0:16:04 | the |
|---|
| 0:16:05 | do you okay yeah i'll to for the simulations right yes |
|---|
| 0:16:07 | okay |
|---|
| 0:16:08 | more comments |
|---|
| 0:16:09 | quest |
|---|
| 0:16:11 | so do you expect this in a performance if you have and the noise uh a non stationary noise so |
|---|
| 0:16:16 | is like the windows or man in exact or |
|---|
| 0:16:24 | you expect a similar performance then |
|---|
| 0:16:27 | one |
|---|
| 0:16:29 | but uh what kind of a do you expect if you have a non-stationary noise or is like you are |
|---|
| 0:16:35 | the windows close the windows |
|---|
| 0:16:36 | i i engine noise exactly |
|---|
| 0:16:40 | two |
|---|
| 0:16:41 | oh |
|---|
| 0:16:42 | i se and do you expect a and that performance or do you a a change algorithm if you have |
|---|
| 0:16:47 | a |
|---|
| 0:16:47 | and noise that it it is just had a yes |
|---|
| 0:16:50 | i i i expect this similar performance |
|---|
| 0:16:53 | or or or all kinds of noise |
|---|
| 0:16:55 | okay |
|---|
| 0:16:56 | okay |
|---|