0:00:13i will
0:00:13show the also has a three minute warning
0:00:16okay uh i can just and the fifty minutes presentations not tick
0:00:20oh it's about exist display
0:00:22yeah yeah hi everybody have time to to be yeah um uh for the morning not to use a scroll
0:00:28do but
0:00:29thank you
0:00:30speak to yeah function that says
0:00:32just as shown
0:00:32so a the mother used mission
0:00:37yeah yeah that's use a level of used
0:00:40uh it was not some a question that it was a a a a a a the question of a
0:00:44review of a of uh this paper
0:00:47and i even if may looks like a bunch question
0:00:54is is the worst
0:00:57use one H
0:01:01so i we just take the first minutes you know to
0:01:04to give you some context and try to send are uh answer this question
0:01:10so the voice production is mainly made of the three components of the glottal source
0:01:15made uh generated of the larynx level of the uh the vocal tract filter are on finite the radiation
0:01:20of the most level
0:01:22and then
0:01:23leading of the vocal a use
0:01:26a propagated with form
0:01:28i i can be perceived from someone i
0:01:30my feel of research is the voice transmission
0:01:34the question is or to transform a probably with form
0:01:38you know that to make it perceived differently in of
0:01:41voice qualities in terms of timber a in terms of
0:01:44uh any other uh perceived elements of the voice
0:01:49and the one uh uh i D
0:01:52split of the propagated located where for
0:01:54in order to retrieve a
0:01:56uh the elements of the fun and of an elements of the voice
0:01:59uh and in any other the all sort of a bit up to some the addition
0:02:05i we use it in order to do this
0:02:07uh yeah uh it is necessary to in but as the voice production which is
0:02:11what a tricky problem
0:02:14and one solution
0:02:15if we need to invert uh uh
0:02:18a search problem
0:02:19uh is to add some constraints and
0:02:22uh for of this and the and models
0:02:24are very useful
0:02:26because the that described an i to go a waveform on add some
0:02:30constrained for these and
0:02:32so the question is now
0:02:34how to estimate the shape by does
0:02:36or for both of model
0:02:40so first i will briefly describe the model
0:02:42or the rows production are used
0:02:45as i say that there is a but also has of a trip fit on addition duration the but source
0:02:49is made of
0:02:50and a shape
0:02:52in the time domain
0:02:53this should as
0:02:54obviously a time position
0:02:57and then we assume that is shape is periodic
0:02:59then uh we assume that the vocal tract is minimum phase
0:03:03or the zeros of these that from
0:03:05it's is is that
0:03:06transform uh lies
0:03:08in side the unit circle
0:03:10and a fun i you we assume that the relation can be simply
0:03:14my of the using a a a time a simple time to do the derivative of
0:03:18which is in frequency domain uh G W
0:03:22i have had the find and then we use a a a each uh representation
0:03:27uh it may that's
0:03:29also the and
0:03:33a monies
0:03:34is simply uh defined as
0:03:37so you will find the now a a a a a phase component
0:03:41as as a complex six L also
0:03:44C and uh a uh it's or addition is simply we use
0:03:49uh G H
0:03:53so the proposed was in this paper
0:03:56is the function of phase distortion what is that
0:04:00uh yeah I Ds
0:04:03it describe
0:04:04you see you know a D question on describing the
0:04:07do the function of phase distortion
0:04:09and uh of the first I used to remove
0:04:12and a medium phase components or from the main argument
0:04:17using a a a a a a a it's uh um me "'em" phrase weight values editions through the real
0:04:29this way of for yeah removing a new face uh real ready
0:04:33for example uh uh to remove the contribution in the face
0:04:37uh of the vocal tract fit uh which could be present in the argument X H
0:04:42then we use the seven or the difference about brought or in order to remove
0:04:46and we now phase component remaining in a uh this
0:04:50and division
0:04:51and finally we use a and the two difference a parameter
0:04:55you know to obtain a a a similar addition to the group do
0:04:59we each is a a a a a meaningful in terms of
0:05:01phase distortion
0:05:05so that to the uh
0:05:08or of those for a function of discussion maybe you at about uh thirty G from but that model
0:05:14which uh described
0:05:16uh the shape
0:05:17of of the a a a a a lot of buttons
0:05:20and we can use uh simple powerpoint use of this but the model
0:05:24a using the transformed the lead you can from a model
0:05:27and this problem to the of the which is used is a a a all the
0:05:32part of the is more a uh of the pose
0:05:35is uh looks more like any buttons
0:05:37and he this bomb to is beak it more looks like a sign is i
0:05:42a but the right
0:05:44a about on of uh
0:05:46was slide you would see a um a plot
0:05:49which describe a function of phase distortion for are on of the first three harmonics
0:05:55uh with respect to uh its shape parameter uh
0:06:00this was function or a a uh in that i don't
0:06:03a many uh
0:06:05elements and there are only related to the shape
0:06:08the i the and on the time position times to this end and all phase difference
0:06:13the and the but i'm to the amplitude things
0:06:15as an an addition by you mean "'em" phase very efficient
0:06:18i and the on to the duration of the buttons
0:06:22uh uh we uh walk
0:06:24on a uh a and money model
0:06:26and find the are independent
0:06:28or the new face call component
0:06:31so those function and the you related to as the uh a shape of the buttons
0:06:40and application is i those function of a distortion is the estimation of a of V R D problem at
0:06:46a jack can from model
0:06:49and in a lot to do this
0:06:51i i will be describe a uh they're could an which is used
0:06:55uh for of these
0:06:56uh estimation which is
0:06:58the great in a phase musician
0:07:01as been a really use
0:07:02a a you know that to the done line uh instant of
0:07:05see if you can X station
0:07:08or first we use the convolutive residual a which is is the division spectral domain
0:07:12or or was the observed spectrum by model
0:07:17then a uh we can say that if the model correspond to zap some signal
0:07:22it can that the comet
0:07:23quality what do that a is equal to one
0:07:26but that's i can see is which you mean that
0:07:28it's amplitude spectrum is equal to one a and needs space spectrum
0:07:32is equal to zero
0:07:34so the idea of great and phase musician
0:07:38to ensure that
0:07:39the amplitude spectrum of the convolutive residual is constant
0:07:43and then a the simple idea is to minimize
0:07:45a face but some
0:07:47you know the
0:07:48to feet
0:07:49the parameter or to the observe scene
0:07:54so we can see that
0:07:55uh a given a of the voice production model as described before
0:07:59no uh the linear phase but the shape of the glottal pulse the vocal tract it under addition
0:08:06and a
0:08:07the function of
0:08:08phase discussion we propose
0:08:10we can see that as functions can be applied to a a a a the convert you raise your are
0:08:14we know
0:08:15scrap the
0:08:16the question but
0:08:17you can just see of the file and uh for the last line
0:08:20but that you can see a representation of the function of its station
0:08:25by means of C
0:08:27division by the you and phase realisation of the numerator or
0:08:31with a an additional in a phrase and this
0:08:34no phase is we be removed
0:08:36thanks to this this and uh all the phase different
0:08:40such as a we can um minimize and the phase spectrum
0:08:44or of the convolutive residual
0:08:47by minimizing is this
0:08:48um L function which is done and the speed is square for mean squared phase using in so one the
0:08:54phase difference
0:08:55which is simply
0:08:56the square of
0:08:58uh a function of phase distortion
0:09:01the types of seeing our by uh the lf model
0:09:07yeah is some example of and those uh all functions
0:09:11we uh three synthetic signal i'm as you can see
0:09:14uh those functions are uh
0:09:17uh uh simple uh uh to optimize in terms of
0:09:21finding being the row
0:09:23the global uh a minimum
0:09:25and describe a minimum uh seems to
0:09:28to close spend while to two of these syntactic video which is used
0:09:32in this example
0:09:34there's we use
0:09:35a simple a um brands method you know that to retrieve
0:09:38the block but the global minimum of those functions
0:09:43in terms of evolution of these uh make that's
0:09:47to a simple example the
0:09:49well i'm about
0:09:50represent a and requiring a problem of of rose
0:09:54going from a right about right voice
0:09:56to a tense voice
0:09:58um in
0:09:59a you can see and um
0:10:02the yeah ad you and or which is a predicted from the Z to the graphic see no
0:10:08and uh
0:10:09a well you can see uh the a deep part two which is estimated through you you D square
0:10:14based made
0:10:16obviously there is a big yeah has even if at the shape of a use uh
0:10:21estimates modes are quite similar
0:10:23there is a big B be could
0:10:25the link
0:10:26the link between uh the work of for motion the make an mechanical of motions
0:10:31and uh the battle per season not
0:10:33is not to use so we can only observe so colour relation
0:10:38and below low you can see in figure
0:10:40a a a a a small segment of speech
0:10:42with writing uh estimate of our department
0:10:51more than a few examples
0:10:53here is an evaluation
0:10:55um made from uh data basis using a game addict or blood the graphic signal for
0:11:01the open question ease of predicted from the uh are D shape part of the and
0:11:07the idea of a open question is predicted from the
0:11:11click or of got sick and
0:11:13and the figure out just below show we use a standard deviation of these open cushion
0:11:18you know for for different that the base
0:11:21and a as you can see a the proposed was method
0:11:24seems to have a on the
0:11:26a state of the main state of the also made a
0:11:32so yeah
0:11:33just as a mean to our conclusion
0:11:36we propose a function of phase distortion
0:11:38which are really did and mean to the shape of the problem birds
0:11:42and as just function can be used
0:11:44so to estimate a a should parameters of local models
0:11:50thank thank for the action
0:11:56thank you
0:11:57and to yes
0:11:58a lot of time for question
0:12:09i start the discussion with
0:12:12in it used uh you model by decomposing uh the total
0:12:17transfer function the product of the glottal model the vocal tract and radiation model and you mate the standard assumption
0:12:22about the vocal tract and radiation
0:12:24in terms of as of their face function being minimum phase so be just the tree a make a term
0:12:30that's that
0:12:31assumption faq
0:12:33you of further analysis at all
0:12:35because you set funds you independent of the minimum of component
0:12:41uh what if there are to track not but not be minimum face in the end
0:12:45yes obviously use these and make use of a track is fine not mean will phase it will uh a
0:12:49lot in found to the the result of uh the
0:12:53of the
0:12:54the estimate of the of the shape parameters
0:12:57and that the same at
0:12:59a are some shot about the vocal tract filter a is quite
0:13:02is not that
0:13:03a a strong and it seemed that
0:13:05uh uh sit assuming that is me more phase is quite
0:13:13a a a a a a good assumption
0:13:15but uh uh i i had made that's for the addition of the mouse a level uh uh it will
0:13:20been may be necessary to make some more about
0:13:23improving the model of the of the models are addition
0:13:28and put that explain the differences you see between us to mets send
0:13:32the class uh cut the graphic estimate
0:13:38and i didn't because
0:13:40um you have to hmmm and to remind that uh the egg that a click
0:13:45electroglottographic signals are only linked to the motion of the vocal folds
0:13:50well what we are trying to retrieve a is an estimate of
0:13:55an approximation of the glottal flow
0:13:57we and and the acoustic be could between the two is uh for form of views
0:14:02not only now and so we can not only observe a correlation
0:14:06so that's why there is a so much difference
0:14:10yes piece
0:14:11because there
0:14:17and that would be your evaluation you compare against the i i i i a i have method this
0:14:22you don't actually you don't say boss
0:14:25privatisation that's of was used to extract a Q
0:14:28which i think might be chris go
0:14:31can you can use a but last method used
0:14:33after using D i i have that it's just this is a a a a a uh i i i
0:14:37if is first to get it to estimate the plot of for on not to estimate part of more than
0:14:43so uh the I is for to estimate the of for and then to fate
0:14:48a a a a a a a a model on some pulses
0:14:50uh a on the estimated uh a lot of rule
0:14:53using a a uh
0:14:55gradient descent or any you know those uh made
0:14:58we can use
0:15:02just just as a from from my screen's the at that the method jeez
0:15:06"'cause" it
0:15:07it it
0:15:08it the make the method use can to different methods use very different a key value
0:15:13so have
0:15:14yeah just
0:15:16thank you
0:15:22you have more questions
0:15:29this is not the case let's say thank the speaker again