Player is loading...
ESR10 : Acoustic features to support the perceptual evaluation of accent production in dysarthric speech
Viviana Mendoza Ramos
Thursday, Sept. 5, 2019 · 11:39 a.m. · 08m 58s
Embed
Transcriptions
Note: this content has been automatically generated.
00:00:01
thank you really a okay with keep that because
00:00:04
they will not to say again there where he um
00:00:10
that was the topic would present in the conference and the at the conference
00:00:15
yeah yeah it was for acoustic features to support the perceptual evaluation
00:00:20
of accent reduction in the set to speech it was a him
00:00:26
it was a mm i work from the
00:00:29
department where we could uh do some are different
00:00:35
uh sorry oh yeah you know
00:00:40
that the from previous to the other slide you'll see
00:00:47
they're not the other one yeah that are um from previous is that is we know that a
00:00:55
in action production that is a a a correlation with
00:01:00
their changes in fundamental frequency and intensity but normally it's speakers
00:01:05
but for this actually is huge they use the combination ah only having
00:01:10
stuff then so so well or the or the accent reduction in sentences cell
00:01:16
uh but we do some here how these differences between accented an an accented
00:01:22
syllables are objectively it's red and how
00:01:25
this changes in fundamental frequency intensity and duration
00:01:29
i related with the president's statement oh yeah example
00:01:32
creation and also with the entire segment was oh okay
00:01:38
those went to call them a search and the next presentation next is like oh well
00:01:48
and yet to identify their eleven features of sentence accent reduction in dutch
00:01:54
related to fundamental frequency intensity and duration within this eleven
00:01:59
in contrast with the previously level i'm also in contrast with the
00:02:03
entire sent and other window classification between x. and pick him accented syllables
00:02:09
accent reduction past with with which is an important eh and
00:02:14
oh yeah aspect of rehabilitation at all he said just severity apples
00:02:19
and to objectively demonstrate the similarities and differences between at the speakers
00:02:24
and eh speakers with this side yeah you accent reduction
00:02:30
the next next slide the and we would get the uh we used
00:02:35
to him this examples of the batteries corpus which was the battle for
00:02:42
rest of the research we it's pretty healthy speakers have fifty
00:02:46
with the set yeah with different severity levels and yeah yeah
00:02:51
with to present this same present and squirt it which is a
00:02:54
subject and that don't show that objective analysis for sentence as in production
00:03:01
and the first uh is the signal is analysed by the contour algorithm so
00:03:06
that takes the slap the nucleus and to extract the acoustic features that together
00:03:12
with the perceptual evaluation and the input for reading and discriminant analysis
00:03:18
in order to uh to that their mind the
00:03:20
most relevant acoustic features and here the next slide
00:03:28
this is an overview off all the features we got relate with algorithm
00:03:33
related not only eh with this syllables
00:03:37
in marin parameters but also with and
00:03:40
in contrast with the previous eleven and in contrast with the sentence all them
00:03:45
are related to fundamental frequency intensity and
00:03:49
duration measurements and the next is like
00:03:54
after the use in the l. d. a. we get to extract the mustard relevant features
00:04:01
that corrected ice axe in production and that life isn't egress the contribution of each valuable
00:04:08
for accent that they should uh and yeah yeah that that it would
00:04:13
be the distance like would it uh it gives like an idea at
00:04:17
that some common features i use score ball robots but it seems that
00:04:23
is because with exactly yeah use different
00:04:26
compensatory mechanisms and they combined merrick strategist
00:04:30
the prettiest antenna accent that a normal the m.
00:04:34
m. o. s. speakers and the next is light
00:04:39
eh with the selected features we got
00:04:41
an accurate classification between accented and accented syllable
00:04:47
what about like while the the office species we
00:04:51
could see that for the stack trachea uh speakers
00:04:55
we get the eighty two percent of incorporate classification
00:04:59
of on yeah yeah accented syllables and as summarising and
00:05:04
like a brief to conclusion makes it like a we would like to say that they have strong but
00:05:11
the strongest predicted ensures i i've been i throw frequency and sitting on the racial but not only
00:05:17
we think that i am still uh which is eh where the accent is located
00:05:22
but also in contrast to the previous eleven and with the
00:05:26
whole sentence and the limited set of acoustic feature is provided
00:05:30
to correct the rise these eh this is his task and the next is light
00:05:38
is just a suggestion that we would like to use this a mm just
00:05:43
features in the development often assessment tools for database that
00:05:47
again yeah yeah is it or object is uh evaluation oh
00:05:52
no this is a this therapy and bake it will be
00:05:57
a helpful because they yeah yeah you could give a better
00:06:01
it did appear metals and approach for
00:06:04
and personalised exercise for each k. page
00:06:08
and in other than x. is like oh it's in other order of information i
00:06:15
would like to say that or the day before a word about the development and
00:06:20
evaluation of the people articulation database i yeah we already has a new thing there
00:06:26
recorded from patients we decide yeah that were on their articulation training
00:06:32
and hit them beta was already perceptually evaluate it and it's currently
00:06:38
eh again i'm not a right and a detailed description
00:06:43
an information protecting the articulation nearer it's been a s.
00:06:49
yeah i'm not take it like you can steal a an example there
00:06:53
we get the eh we're annotate and the the tack that well or that
00:06:57
i got i mean this so that if it was a a directly uh it
00:07:04
and percy if or if it was a
00:07:06
substitution in case it was a substitution we
00:07:10
the the annotation uh it off which here
00:07:14
what was the what each is an ad here
00:07:18
and in cases that don't get on it was perceived but with distortion
00:07:23
we measure with the visual and mel scale
00:07:26
that the distort the distortion level of the funding
00:07:31
and the i would the utterance line i would
00:07:35
like to mention that i already went to to
00:07:40
great to the uh to the basic idea of where lenin and idiot eh easy to do
00:07:46
like yeah well my comment and i did yeah
00:07:51
sure stay yeah hash and it to an inside
00:07:57
in different models that a machine learning models with that at worst and
00:08:02
i would like just this and other players approach off the bus them
00:08:10
stress cults we've got you with a it and university
00:08:15
i would get that that number because it shows or
00:08:19
or well yeah it for just this reason with train again
00:08:26
yeah yeah then and read with the c. d. and database
00:08:29
and the models eh yeah just with the rate of speech
00:08:35
and eh the there s. all is the
00:08:39
good but we would like to rub it in the future
00:08:43
yeah i think that was a short summary of or what they have done this year
00:08:52
every well for your attention and i
Conference Program
Storage and management of sensitive speech data
Henk van den Heuvel
Sept. 4, 2019 · 11:51 a.m.
ESR01 : Speech based automatic dementia detection in a home environment
Yilin Pan
Sept. 4, 2019 · 2:14 p.m.
ESR03 : Interpretable speech pathology detection
Julian Fritsch
Sept. 4, 2019 · 2:30 p.m.
ESR04 : Acoustic speech based monitoring for telehealth and predictive analytics
V. Srikanth Nallanthighal
Sept. 4, 2019 · 2:54 p.m.
160 views
ESR07 : Modeling progression of patients with neurodegenerative disorders
Camilio Vásquez
Sept. 4, 2019 · 4:05 p.m.
ESR08 : Developing measurment procedure of pathological speech intelligibility
Wei Xue
Sept. 4, 2019 · 4:31 p.m.
ESR09 : Clinical relevance of intelligibility mesures
Pommée Timothy
Sept. 4, 2019 · 4:49 p.m.
Big Data with Health Data
Sébastien Déjean
Sept. 5, 2019 · 9:20 a.m.
Small Data, Number of features vs. number of observations
Olivier Crouzet
Sept. 5, 2019 · 9:47 a.m.
ESR11 : First year review
Bence Halpern
Sept. 5, 2019 · 11:20 a.m.
ESR10 : Acoustic features to support the perceptual evaluation of accent production in dysarthric speech
Viviana Mendoza Ramos
Sept. 5, 2019 · 11:39 a.m.
109 views
ESR12 : Continuous speech recognition for people with dysarthria
Zhengjun Yue
Sept. 5, 2019 · 11:50 a.m.
153 views
Management of health data in an industrial company
V. Srikanth Nallanthighal
Sept. 5, 2019 · 1:57 p.m.
146 views
ESR02 : Automatic detection and classification of pathological speech conditiond based on emotion expression
Zhao Ren
Sept. 5, 2019 · 4:34 p.m.
ESR13 : Bridging the gap between typical and pathological speech recognition
Enno Hermann
Sept. 5, 2019 · 4:56 p.m.
ESR14 : Developing speech therapy games for children with speech disorder
Thomas Rolland
Sept. 5, 2019 · 5:16 p.m.
ESR15 : Automatic assessment of speech in cochlear implant (CI) users
Tomás Arias-Vergara
Sept. 5, 2019 · 5:32 p.m.
Recommended talks
Dysarthria
Marc de Bodt, Antwerp
Sept. 24, 2018 · 9:45 a.m.
302 views