0:00:14a mind then these heroes you guys too
0:00:16yeah and the whole
0:00:18also uh i
0:00:19to see something like that
0:00:22we still go you best about a week right on the technology to file
0:00:25my project
0:00:26uh tied to the title my presentational
0:00:30hayes class case um by waiting for specific case you component well but signal three commode a much
0:00:40the application a well what is that is the brain computer interface
0:00:44oh what is that it to reconstruct
0:00:47extract the future it's also sit rank condition
0:00:51wrong brain signals
0:00:53to exploit the bci
0:00:55as sort of features correspond to break conditional
0:00:59such as
0:01:00uh as as bp
0:01:02uh are four where
0:01:04and can be observed by observational system
0:01:07such as the G
0:01:09and fmri
0:01:11a piece yeah it's kept yeah such speech at and uh saying that the base command
0:01:15i correspond to each plane condition
0:01:19in this study we focus on but uh it's measurement missile
0:01:23and that our future
0:01:24induced but i
0:01:26imaging environmental forty
0:01:29yeah yeah we show uh try your uh
0:01:32what they match the deepest pca
0:01:34yeah we call me might be see i
0:01:37i use them
0:01:38you meant in the movement of the hand
0:01:43set i'm feature
0:01:44can be able seven inning
0:01:46brain signal
0:01:50weighted reports da electricity they sell
0:01:53i might fit for that
0:01:55in this example we use
0:01:57uh some electoral
0:01:59and the system called electoral so it's a photograph E E G
0:02:05uh observed signal uh
0:02:08transform to a device command
0:02:11cms processing step
0:02:13for feature extraction and tone recognition
0:02:21by this system
0:02:24uh but it's
0:02:25speed it uh
0:02:27uh and uh face
0:02:28uh can
0:02:30jen control
0:02:31any device
0:02:33so that for example for it yeah and mascot or a computer display
0:02:42muscle movement
0:02:47yeah we
0:02:49three briefly
0:02:50explaining the future
0:02:53yeah feature to be explicitly by
0:02:56in in maybe see i
0:02:57it is known that uh energy in frequency band called mu reason this desynchronized by
0:03:03means a movement
0:03:04uh well well
0:03:06that is synchronise the location
0:03:08in the brain depends on imagine but
0:03:12and therefore
0:03:13if we know a change over the energy in something frequency band
0:03:17and the special locational we it's energy
0:03:21uh we can recognise
0:03:22imagine the part of uh body movement from easy signal
0:03:28uh in it might be see i
0:03:32the most bizarre buttons
0:03:34csp missile
0:03:35uh is by the use
0:03:37and especially
0:03:39uh to grasp
0:03:42but where might be vision
0:03:45and yeah we
0:03:46a a and it is
0:03:47out out we show that the line of the but it wouldn't it
0:03:54a the red uh it is signal
0:03:58a was out by but it's and it with them of the system
0:04:01and uh signal
0:04:03but post filter
0:04:04and you know
0:04:05the plus on the but this pond post filter
0:04:08a a set of the say but this that it hurt
0:04:11uh the
0:04:13the case you bounce
0:04:14all the separate is that it had
0:04:16uh contains a a a pretty K C bound
0:04:20a codes
0:04:21um you and the it
0:04:24uh a that signal
0:04:26oh wait it
0:04:27by W
0:04:28and this is a special weight
0:04:30and uh we get that doing a competition
0:04:34like this
0:04:38is S P
0:04:39as as the weight if this has its from so weights W decided from running it
0:04:47the by this of the my this program
0:04:52uh this stuff in this of program
0:04:54take that you expect a shell with a body as of a special so weight
0:04:58it signal
0:04:59a but on the T they don't need to class C
0:05:02is minimize
0:05:03where she is that of sectors about it
0:05:06and this lies see a set of a
0:05:08data but on to belonging don't to class C
0:05:11and the E X P expectation of a all by C D
0:05:15grass K shall we can use the body as
0:05:17oh but extract signal
0:05:19uh as each about
0:05:23oh uh they a P need band post filtering
0:05:26uh as processing
0:05:28a hyper
0:05:29uh a like this
0:05:30i have a doubt them a three can see bound depends
0:05:33on on um it's "'em" at the member i matt
0:05:36and toss up
0:05:38we won't
0:05:39that just fitted that take is direct component that most D contain
0:05:44to just the eight to to a
0:05:46what the image
0:05:49uh some muscle
0:05:50have been proposed to
0:05:52design the filter
0:05:54uh a fast super missile
0:05:58we set
0:05:59this uh
0:06:00we set is uh was to to every button out of the a candidate
0:06:04uh uh running that the by cross
0:06:06and this to D so
0:06:10E to file
0:06:11designed uh i hope to my for that mike is endings
0:06:15a C P
0:06:18i well yes this P cannot provide that a complete get three cases selected
0:06:22and the C S S as P
0:06:24and introduced thus the of the coffee C N or or or a a a i on filter
0:06:30a high bar
0:06:32you know the might this
0:06:35oh to my the complex
0:06:36and time consuming
0:06:38and the specs just V is design
0:06:41the this be designs uh
0:06:44in for casing domain
0:06:46i have a that's an get T that that of to my addition can
0:06:52so in this paper we propose a use improve missile to design the specific case see filter
0:06:59i as fast
0:07:00i fast fast idea
0:07:01we uh the in do the cost function just P
0:07:05i put that's that the rates for frequency component
0:07:09and do we look to you
0:07:11uh to optimize
0:07:12that was function
0:07:15the proposed cost functional
0:07:18uh there but then need to at up the my patient for both
0:07:21and specific filter
0:07:23and a free case for freedom
0:07:24and then zack cost function can but because we use
0:07:28well these the cost
0:07:32a a mobile a top to my addition supplements romance
0:07:35in a the in team up to my addition
0:07:37uh did use to denies the a but you program
0:07:44so we
0:07:46yeah i expanding
0:07:47to will propose missile
0:07:50for at understanding under standing we you don't to late uh
0:07:54and a mental a signal
0:07:56uh like
0:07:57but uh of the blocks
0:07:58like this
0:08:01the crime of a
0:08:04oh by X is then not channel
0:08:06and uh a low uh a low you that time
0:08:12oh it's pretty demon only
0:08:14uh processing
0:08:16we define the uh
0:08:19the spectrum all move that windowed signal
0:08:21a fast we
0:08:23defined that window signal
0:08:25add to the X
0:08:26and the
0:08:29to the X to it's a transform the spectral
0:08:33uh to the Y
0:08:34a by the speed of on a school
0:08:41like is
0:08:42uh we define in then you feature value
0:08:45using a spectrum
0:08:48and this that yes
0:08:49a a rate that
0:08:51how to act to W on the eight
0:08:54because uh is uh
0:09:00uh the current Y
0:09:02uh uh three guys component
0:09:04and the
0:09:05little or why
0:09:08the case C component
0:09:09uh that we do
0:09:11that's the so wait
0:09:12and the H like the spectral weight
0:09:14but it at is uh
0:09:16because for
0:09:19in a "'cause" we will show how to find of W
0:09:22and eight
0:09:26uh we designed this of to mine is a program
0:09:29uh for four was weight picked up by extending this just cost
0:09:36in this
0:09:36to my that program
0:09:38uh uh but a C on and it's yeah this a
0:09:41decide such that station all with future
0:09:44which each i that
0:09:46uh they don't the class C it to minimize the
0:09:49in nineteen data
0:09:51and and as a a normal vision or all sampled
0:09:55we see that i
0:09:57of not cross rubber
0:10:02so we sort of
0:10:04the uh
0:10:06however it did this got to estimate and the city a to my
0:10:10uh both W and eight
0:10:12so we they by
0:10:13to that problem we into two stop program
0:10:17uh the program
0:10:19to optimize
0:10:20W while fixing eight
0:10:23and the to optimize H where thing that way
0:10:28not think that can be transformed to the a
0:10:31G is for
0:10:32we can define an to supplement
0:10:35oh yeah O R is a whole body has what the ticks
0:10:38i don't to
0:10:39and a channel component
0:10:41and the Q
0:10:43but it also co but it's not the ticks alone two
0:10:46uh uh
0:10:48and i two because component
0:10:50it isn't a
0:10:51well known
0:10:52that's uh
0:10:54D supplements can be deduced to by this to generalise at "'em" body problem
0:10:59i by proof
0:11:02and this is somebody of that
0:11:04of the bat um process it but C
0:11:07as the flat
0:11:08is eyes
0:11:09the eight
0:11:10and the step to
0:11:12of them that but you by serving being program well
0:11:16and the step C D
0:11:17of is eight
0:11:18by a serving supper made to
0:11:22so we did P
0:11:23as step to to and the steps three
0:11:25and do post function it's
0:11:32oh mobile we will do so
0:11:35uh reports
0:11:37you also propose that a
0:11:39classification do
0:11:41uh a cross lie
0:11:43and the be did that a
0:11:47yeah i of the one you is uh
0:11:49you these different to affiliate the eight between classes
0:11:55we use that one on the H mount H to
0:11:59in this
0:12:01and ms of to D is calm of filter
0:12:05being or thought or seeds
0:12:09fast fast we find uh
0:12:11W the
0:12:12and at G
0:12:13in the we had to outside "'cause" rubber
0:12:17after after thus
0:12:18be sort of a sub one with a T G
0:12:22in other that class
0:12:25and G but
0:12:26and the we get stuff you but
0:12:30in this dismissal
0:12:34we use
0:12:35W one on the bridge to and the a G
0:12:42uh we
0:12:46uh to off on the uh
0:12:48propose missile
0:12:50but this dataset set
0:12:52uh name data set four we have from bci competition three
0:12:57this data sets contains that's a C G signals
0:13:00cross the cross is are right hand and right foot motor imagery cross
0:13:05and five subjects but at
0:13:08and the easy signal a a what it to be
0:13:10well hundred eight vector all
0:13:13and sampling for K C
0:13:15i is a hundred ten
0:13:19the subject
0:13:20but home of the
0:13:22uh task
0:13:25two hundred eighty times
0:13:27so the number of out of in each class is more hundred forty
0:13:31and do a shell beach prior three point five second uh that's it tough sing
0:13:36that it's thus
0:13:37signal ranks
0:13:40so hundred
0:13:41fifty some
0:13:45we can be as uh
0:13:47problem is all we
0:13:50uh a few text task meant to texas some feature extraction missile
0:13:54uh says be one
0:13:56use use a bunch was with uh with the past bound of separate is that to have
0:14:00and the P to
0:14:02you'd use the bound was create that when idea optimized by a C B
0:14:08and do you also used
0:14:09a that uh that are designed to by C S S P and the specs
0:14:13S P
0:14:15a a in the in the missile
0:14:17we adapt to simply classification blue
0:14:20at to compare yeah a future value correspond to
0:14:25uh it's class
0:14:28for example you S
0:14:30csp P
0:14:30we use discuss kitchen
0:14:35this a the cross case and is that
0:14:39uh this there was so that
0:14:42uh chris case accuracy
0:14:43uh a given by a five five five by five cross what addition
0:14:50for a subject and each missile
0:14:53you know but eight
0:14:55the proposed mess
0:14:59a to home
0:15:00as a miss
0:15:03and uh and the kissed we show that uh
0:15:06we show the uh a a a filter
0:15:09yeah given by
0:15:11each missile
0:15:13in this case
0:15:17in in this case yeah we use a
0:15:20in this case uh W one to that
0:15:24in each each that i is same
0:15:26bit doing we to that of the one with the because
0:15:29uh i ms of that that we use
0:15:32we said to G
0:15:34we set D other one
0:15:37cross well
0:15:38and we show that class
0:15:40we show uh
0:15:43we show W wanted up W
0:15:45W one day about
0:15:47uh since
0:15:48this right
0:15:49this right
0:15:58that this C D S shows a few to
0:16:01and given by dismissal
0:16:04a a kind of to characteristic of the field that
0:16:07given by this missile
0:16:09the propose a not give the same about of the are you
0:16:14and in subject to yeah
0:16:19and in this
0:16:20uh is subject to a
0:16:22and uh propose missile also have a large you eight
0:16:25in but
0:16:26in band high at uh
0:16:29i manually any optimize to plus one
0:16:32and this
0:16:33uh a high yeah bad looks uh a looks the how many "'cause" or this bound
0:16:40this it is a a a a we show
0:16:44uh to look at a show
0:16:47all this
0:16:48it's away way
0:16:51in subject a a a or miss not came almost same space way
0:16:56a last
0:16:57in use of a just a one
0:16:59uh we
0:17:00we are able serve that difference
0:17:02define an
0:17:03us bits are but that
0:17:04so a different frequency component
0:17:07yeah a different is to but that
0:17:13a good job
0:17:14uh we propose a missile to
0:17:17designed to for channel now and the frequency component using running data
0:17:22that's so show over the way top to by this
0:17:25oh the weight of the my this shown can be obtained at thing so
0:17:29the sort sort of being some
0:17:31and then i say can body problem
0:17:34uh uh can adapt
0:17:36can be tough
0:17:38then the cost function is don't increasing in each iteration
0:17:41the proposed missile
0:17:42see how yet "'cause" case accuracy in what they the best is yet
0:17:47a lot
0:17:53or present like we already developed to the temporal fit that design miss
0:17:58and uh and corpus was a temporal filter
0:18:02design result
0:18:03uh a uh
0:18:04to be present
0:18:07thank Q