0:00:14uh i'm technical technique you in a from the uh yeah i T in japan
0:00:19that are are can can say shame but to feature extraction for three major jocks or something in
0:00:25in this walk
0:00:26we focus on the sense of device or or or three mean jokes or meter
0:00:32i don't such X or a is becoming a a a are sensitive by
0:00:36since it is and very in more so
0:00:39a smart phone such as i phones
0:00:42a i mean you of you might have such yeah smart phones i think
0:00:46and and uh
0:00:48and the same provocations use in there are such
0:00:52a signal meter is a screen rotation
0:00:55by detecting the orientation
0:00:57all utterances of the by source model for uh like these
0:01:03um this is a very convenient and useful for users browsing
0:01:08but uh i saying that beyond size seen propagation
0:01:13the three main jocks zero meter in smart phones
0:01:16has a potential
0:01:18to provide more source
0:01:19so katie pairs application
0:01:22or directly to the human action
0:01:25for examples
0:01:26uh action recognition for identifying data sparseness situations
0:01:31on that path somewhat identification such as by kate
0:01:35uh for security and so on
0:01:38i don't by
0:01:39in case oh at recognising human options
0:01:42the problem is that
0:01:44there are three dimensional acceleration signals are rotated
0:01:48according to the device orientation
0:01:52oh falls are a simple obligations and such a screen or stations
0:01:56it is forever used
0:01:59uh for recognizing that human actions
0:02:03it is a big problem
0:02:04he calls
0:02:06the human options are you don't want to festivals orientation
0:02:11so in this walk
0:02:13or or uh our core is to propose an efficient method for extracting rotation invariant features
0:02:20for three-dimensional acceleration signals
0:02:23without any preprocessing for recognizing human actions
0:02:27independently of the device orientation
0:02:31uh you know charlie T
0:02:33uh the rent that the bias is
0:02:35device or an additional is changed
0:02:37uh there are also acceleration signals also rotate it
0:02:41even for the same option
0:02:44so in this case
0:02:46and and tsar rotation invariance means that we can't ten the same features from the same options
0:02:53no what the how that device or oriented
0:02:56the proposed missile is based on that frequency characteristics
0:03:00proof do used by it's are fourier transform
0:03:03we can see there are the corporation
0:03:06a long all frequencies
0:03:09not sure tend to a or doing prospects on how right
0:03:14oh of us let me uh a views the to rate to walks
0:03:19fast ms or is based on the a simple statistics features
0:03:24these so
0:03:25a fast is to me is uh a vector or you know to the by the
0:03:30uh by applying the moving average to the of south
0:03:33a a and signal
0:03:35and the and there uh
0:03:37uh four kinds of statistics
0:03:39uh such as a mean and uh
0:03:42a standard deviations and uh are mean on mark's not a power created from
0:03:47three types of and uh a feature body
0:03:51uh such as uh it's normal of the observation big so and for that um course that between the
0:03:59oh so some back to a on the estimated graph depicted Z
0:04:03and uh uh second missile
0:04:06is based on the way it's features
0:04:09this look or so of uh how fast is uh up to the high so
0:04:14a C eight to zero so a source some signals
0:04:18um then uh
0:04:20a us at the level all way break a car create from the
0:04:25from those uh principal at three priest can point
0:04:29in these two missiles
0:04:31and uh
0:04:31the start is got preprocessing
0:04:34a you know these by a a bit strings
0:04:36uh and probably
0:04:38in a to compensate so
0:04:40and as you know station
0:04:42uh but
0:04:43such pre-processing required that
0:04:46and trying the of a signal data for statistical a stable words that
0:04:50so in this work we propose a novel
0:04:53feature extraction of some um sort which is in can three wrote at embedded
0:05:01and in the proposed missile so we focus on the frequency you relations which is
0:05:06uh useful to effectively characterise a few more options
0:05:10be calls the us assume a options are mainly consisting of a a
0:05:16cyclic a P all the motion
0:05:18such as a walking running on so on
0:05:22so two weeks of the frequency motion
0:05:25we have lies is that if if T actually short on if is T
0:05:30to to their of the of the house
0:05:32oh are some signal
0:05:34and uh we use uh for
0:05:36so you know
0:05:38two describes the proposed missile
0:05:41do you know that are of obtained
0:05:43three main jocks or some signal
0:05:47in uh X Y and Z axes
0:05:50if if all me a in the case or
0:05:52so we the main general fourier if
0:05:54okay frequency see features
0:05:56also a a C all make
0:05:59and the in this case it up to a is a fruit of T two there
0:06:03and as read them engines uh in uh i X Y and the respective not and a joint three
0:06:11so was a fine oh uh long he's uh that's F of which in the case of free feature metrics
0:06:17of so
0:06:18a T is three D a frequency features F
0:06:23oh this rights
0:06:24shows a a core of the proposed missile still
0:06:28and uh
0:06:29we can is the correlation metrics of as the fourier features F
0:06:34which is a given by this
0:06:38that that this four
0:06:40and uh
0:06:42and uh what don't here shows the process to construct a correlation metrics are which is and by and semantic
0:06:49matt six
0:06:50and here uh
0:06:52starting in also complex
0:06:54conjugate kate sports
0:06:56so in this course three
0:07:00and that time no am and a corresponding to the old you know you power to because the corporations between
0:07:09identical or or a frequency features as resulting in uh
0:07:15and absolute areas
0:07:17on the other hand
0:07:18uh the uh of time diagonal elements uh
0:07:21cross-correlations between various kinds little frequency features
0:07:26so we can say that
0:07:28this improve call matt three
0:07:30not it contains
0:07:32not only is a
0:07:33all the emperor spectrum but also the uh
0:07:36cross corporation
0:07:37and which trained betas as frequencies
0:07:41the proposed feature are are sort the body is all of a about trying to
0:07:46sorry right hand good components of the same matt are
0:07:50a uh i as you can see this is very simple formulation
0:07:54and the you can easy improvement it
0:07:58uh uh the problem
0:07:59a a computational cost is also a little
0:08:04uh we show was the invariance properties is
0:08:08pro involved that proposed missile
0:08:11there are two kinds of in there S
0:08:13uh fast one is the invariance to rotation
0:08:18that is the local main can someone in this
0:08:21in this work
0:08:23and uh
0:08:24and you rotation mark X is denoted by
0:08:27a uh uh as a a
0:08:29and not so that uh this or the semantic sees a a you only needed
0:08:35and uh by using this for the semantics a
0:08:38and is signal on the can so you can't or digits
0:08:42and three features that are describes so that D
0:08:48uh the correlation metrics are all was uh rotate is signal now
0:08:53a is identical to so all of the
0:08:57or or no not to rotate data signal
0:09:01by using these you type deals that also
0:09:03i rotation matrix
0:09:05so propose a feature is
0:09:07uh rotation in there
0:09:10next uh a second one is the in to the a temporal six as that is
0:09:15then try that
0:09:17in this case a we you know so then pride be eight by using the top
0:09:23and by using this time uh and there are a right signal on the can see can do that
0:09:29a features that are
0:09:30describe a that these
0:09:34the of a select bar use all of the corporation board and
0:09:38a point
0:09:38as as that is little or fee propose features
0:09:41is also identical to that all
0:09:44the all reason not do rate signal
0:09:48so our proposed feature is also in there and to ten proceed
0:09:53uh such ten perceive is
0:09:56a a close by R
0:09:59short down if T
0:10:00so this
0:10:02and a in there as two ten it is uh a a useful in prague
0:10:11and we show so that we need to reduce there a post processing in do the proposed missile
0:10:16so of that of the original signals uh uh actually role and transpose by it is uh
0:10:22transfer function do to bear as min here such a causing
0:10:26which is depending on how to are the pisces score
0:10:31yeah action eight
0:10:34is a action is not a directory of self but some who estimate estimated
0:10:39and such a scroll thing
0:10:41a if was transfer function is denoted by B
0:10:45and there's the of fourier of features of the uh uh of sub two signals is also trust full
0:10:52by using this
0:10:54at transfer phone function B that the
0:10:57and so uh a therefore for a we so a little all as a and propose a features
0:11:03to two in you you a price
0:11:05a fixed codes by the transfer function be
0:11:09the little little of the proposed feature is
0:11:12decomposed into these
0:11:14you is four
0:11:17in these phones uh the effects of the transfer function is its represent two
0:11:23that this bias
0:11:25and this is the same much patient not seen the some features full audio processing
0:11:32therefore are the in your must made on is such as that feature this unit that now is could start
0:11:38this guy can out
0:11:40that those by is
0:11:44so uh this is their experiments that were conducted in this walk
0:11:49and of the task or use a to identify a few more or human gait
0:11:54with that any control for this experiment
0:11:58we you only ask the uh fifty eight are sounds
0:12:02to three you walk in daily life
0:12:04a with holding i've phones
0:12:06in their hearts
0:12:09we we this C is uh are quite challenging task for i in to find that a few more
0:12:16the experiment or set there please uh uh that's for roles
0:12:19we have to lies are seventy frames running windows for short time we for T with
0:12:25five T five frames tape side
0:12:28and uh and uh we and the except these as from there
0:12:33lining windows
0:12:35are classified into fifty eight to much cross um
0:12:40a by using a uh a nearest the means in their space of which are this content analysis
0:12:46i'm the oh is that
0:12:48i C is classified
0:12:50by using the mustard to boarding
0:12:52and uh we miss as the accuracy averaged a course the power phones by using C for course patient
0:13:02yeah a fast we shows that performance is an the setting sort of the proposed missile
0:13:09it's of uh three feature has
0:13:11are decomposed into two D C and that is can point
0:13:16uh that these you component
0:13:18is a zero frequency features
0:13:21the is can see that to be that gravity T pick so
0:13:25and so
0:13:26and these
0:13:28for kinds of was sitting that can see
0:13:31uh a C components only a and uh is C plus dc components which is a focal point
0:13:38and uh is C components with a little post processing
0:13:43and uh a C press D second points with a a little cost
0:13:46uh a processing
0:13:49so uh for the uh that are not feature as an which corresponds to that they re prospects and um
0:13:56by taking a low and is there a performance is and if the entry increased
0:14:02and that the best performance is of ten of and uh uh a a a a C of uh K
0:14:07C with a little sitting
0:14:10and the next one is uh a uh of that no features
0:14:16and in this case
0:14:18the best performance is of ten in the second or and a is the press C component
0:14:24but uh in for these an L of that in a few
0:14:28she shows
0:14:29by taking a little can and ends up performance
0:14:32is like three rate it
0:14:35and and up and a reason is that to
0:14:38uh roughly speaking
0:14:40and by taking a little the corporations and B to and there's frequency features P its prey got broken in
0:14:47to are just summation
0:14:49and uh
0:14:50this and the power or is also degraded we think
0:14:55and finally i as a for the for feature
0:14:59and uh average of D
0:15:02or is out
0:15:03and the best performance it's of tend well uh in the fading all its C plus T with a a
0:15:10a a little post processing
0:15:12uh you on show
0:15:14and that we also can cut is uh
0:15:18that a novel features on of that in features in the
0:15:21best sitting
0:15:24the performance is a for are improved
0:15:29and next we shows that a performance comparison to does not to missile
0:15:34and uh uh a as you can see the proposed missile is a significant three or to the other two
0:15:43and these two other men use Z as a preprocessing four
0:15:48uh normalising norm are there is some signals orientation but
0:15:53uh our proposed missile
0:15:54and to not use
0:15:56that's uh uh uh uh a preprocessing
0:16:00and present also
0:16:01the proposed minutes so is uh
0:16:04or so
0:16:05competition a uh a sorry
0:16:08uh the proposed missile
0:16:10and or so
0:16:12yeah a the proposed a have a complete that cost in the proposed missile is also a little compared to
0:16:19to to other missile
0:16:22and uh this is that something that supplement the experiment results uh in there some we did pay paper we
0:16:28only up to the classifier parser
0:16:31this comment comment ice is
0:16:32but in this case we also lies a go as far as of linear know it's fee and to the
0:16:38proposed minutes so propose a feature
0:16:41the performance is almost the same as that of the uh
0:16:45and this kind of analysis based cost it
0:16:48only the a zero point one percent is increased
0:16:53so what the conclusion
0:16:55uh we have proposed a mess so to so the road ensuring bad features from three to means know separation
0:17:02the proposed missile is based on the uh correlation metrics of forty if we can see feature
0:17:09containing a that prospects so in the that when no elements and the cross correlations in of that and i
0:17:16am and
0:17:17and the proposed feature is
0:17:19in parody invariant to rotation
0:17:22i there is a and perceived result any as the whole pretty uh
0:17:29and in the experiments only if you might be shot yeah identifications sort of proposed so these superior to the
0:17:36other two missile
0:17:38and uh
0:17:40and there uh it should be not use that uh the proposed minutes so is a general
0:17:46so it is a go to a break up to to the other tasks using their are there is some
0:17:52signal such as that a segment being on the recognise in there as human option
0:17:57and this is our feature walks
0:17:59i thank you for a