seminar m.tech
TRANSCRIPT
-
7/26/2019 Seminar M.tech
1/49
Review # (or) Seminar #
Speech Processing
By
Name of the Student
Regd Number: 11111111111Course: M.Tech ( CSE )
Deartment of Comuter Sc!ence " Eng!neer!ng.
#nder the$u!dance of
Sr! %acu&ty Name
Des!gnat!on of the %acu&ty
Raghu Institute of Technology
Approved by AICTE(New Delhi) Affiliated to !NT"#$
Da%a&arri(') hee&unipatna& ()
'isa%hapatna& District Andhra *radesh India+
01/30/15
-
7/26/2019 Seminar M.tech
2/49
Objective
Fundamental definitions
hat is speech!
Phonetics and Phonolog"
Speech #ecognition
Speech S"nthesis #esearch areas in speech
-
7/26/2019 Seminar M.tech
3/49
Fundamental $efinitions
-
7/26/2019 Seminar M.tech
4/49
Sound %aves
& sound is simpl" a disturbance of air molecules'%hich radiates out%ard from its source' in %avesof fluctuating air pressure' li(e ripples from astone dropped in a pool)
*he structure of these sound %aves distinguishesone sound from another)
hen sound %aves hit our eardrums' nerve cellsin the inner ear detect the structure of thevibrations' and the" pass this information on to thebrain.
-
7/26/2019 Seminar M.tech
5/49
Fre+uenc" and amplitude of a
%aveA lower amplitude, higher frequency wave:
A higher amplitude, lower frequency wave:
1 cycleof the wave (trough to trough, or peak to peak)
-
7/26/2019 Seminar M.tech
6/49
Pitch and loudness
*he fre+uenc" of a %ave is heard as its
pitch.
*he amplitude of a %ave is heard as its
loudness.
-
7/26/2019 Seminar M.tech
7/49
[ A j u w k I N l e t n n i ]
Spectrograms $ispla" of timeon the ,-a,is' frequencyon the "-a,is' and the
higher-amplitudefre+uenc" regions sho%n as dar(er areas
Spectrogram of 'Are you working late, Nanny?'
-
7/26/2019 Seminar M.tech
8/49
SP..
,hat is it-
.inguistics
*hysiology
Acoustics
-
7/26/2019 Seminar M.tech
9/49
/pea%er .istener
.inguistic
level
*hysiological
level
Acoustic
level
*hysiological
level
.inguistic
level
The /peech Chain (Denes 0 *inson)
-
7/26/2019 Seminar M.tech
10/49
.inguistics
"nits of language+ ,hat are they-
,ords- /yllables- /ounds-
,hat are the individual sounds in language-
*hone&es+ 1ow are they defined-
-
7/26/2019 Seminar M.tech
11/49
*hysiology*his relates to ho% the sounds are produced through neural
and muscular activit")
e set air coming up from the lungsin motion using our
vocal cordsand then %e can channel this air through the
vocal tract using out tongue' lips' etc)
e can classif" the different sounds %e ma(e according to ho%
%e set the air in motion and ho% %e channel the airstreamsthrough the vocal tract)
-
7/26/2019 Seminar M.tech
12/49
Acoustics*his describes the generation and transmission of the sounds)o% air is set in motion)
e generate sound %aves) hat do the" loo( li(e!
-
7/26/2019 Seminar M.tech
13/49
*12NETIC/ AND *12N2.234
*honetics concerns itself with5
*he stud" of the acoustic detail of speech sounds and ho%
the" are articulated
*honology concerns itself with5
onsiders ho% these speech sounds are used %ithin
languages
deals %ith the mechanisms / rules / processes that
underlie / govern these units of speech)
-
7/26/2019 Seminar M.tech
14/49
A..2*12NE
& conte,tual variant of a single phoneme' in a particular phonetic
environment) *he" do not involve a semantic contrast
their distribution is mutuall" e,clusive an allophone cannotoccur %here another can) Predicted/governed b" phonological rules)
For e,ample2 *he p4 sounds in the .nglish %ords pin4 and spin4
are acousticall" different) *he p6 in pin is produced %ith a breath of
air follo%ing it 7aspirated8 %hereas the p6 in spin is not)
-
7/26/2019 Seminar M.tech
15/49
'owels
sounds produced %ith no obstruction to the airstream as it passesthrough the vocal tract)
*here are three main organs of speech involved in changing the
si9e of the air chamber) *hese are
the lips # rounding spreading
the lower 6aw # lowered raised
the tongue # raised flattened brought forward etc+
-
7/26/2019 Seminar M.tech
16/49
Consonants
onsonants are articulated b" restricting the airflo% at some part
of the vocal tract)
*he consonant that is produced is determined b" three factors:
place' manner and voice)
Characteri7ed by three features
18 Place of articulation- ;ilabial'$ental' &lveolar' Palatal'
-
7/26/2019 Seminar M.tech
17/49
*laces of articulation
ilabial;ilabial sounds are those sounds made b" the articulation of the lips against each
each other) .,amples of such sounds in .nglish are the follo%ing2 b6'p6'm6
Dental
$ental sounds are those sounds made b" he articulation of the tip of the tongueto%ards the bac( of the teeth) Such sounds are not present in Standard &merican
.nglish' but in some hicano .nglish dialects and certain ;roo(l"n dialects' the
sounds t6 and d6 are pronounced %ith a dental articulation
Alveolar&lveolar sounds are those sounds made b" the articulation of the tip of the tongue
to%ards the alveolar ridge' the ridge of cartilage behind the teeth) .,amples of such
sounds in .nglish are the follo%ing n6'l6
-
7/26/2019 Seminar M.tech
18/49
anner of articulation
*losive8/top
Plosive sounds are made b" forming a complete obstruction to the flo% of air throughthe mouth and nose)
e,plosion of air causes a sharp noise)
-
7/26/2019 Seminar M.tech
19/49
/yllable
& s"llable is a structural unit of sound that constitutes a se+uence of consonantsand vo%els) At is hierarchicall" composed of three parts2
Onset initial consonant or consonant cluster
?ucleus the vo%el
oda final consonant or consonant cluster
s"llable
onset #ime
oda?ucleus
str
eh n, ths
-
7/26/2019 Seminar M.tech
20/49
-
7/26/2019 Seminar M.tech
21/49
-
7/26/2019 Seminar M.tech
22/49
-
7/26/2019 Seminar M.tech
23/49
-
7/26/2019 Seminar M.tech
24/49
-
7/26/2019 Seminar M.tech
25/49
-
7/26/2019 Seminar M.tech
26/49
-
7/26/2019 Seminar M.tech
27/49
-
7/26/2019 Seminar M.tech
28/49
-
7/26/2019 Seminar M.tech
29/49
-
7/26/2019 Seminar M.tech
30/49
-
7/26/2019 Seminar M.tech
31/49
-
7/26/2019 Seminar M.tech
32/49
-
7/26/2019 Seminar M.tech
33/49
-
7/26/2019 Seminar M.tech
34/49
-
7/26/2019 Seminar M.tech
35/49
.,isting S# s"stems
$ragon ?aturall" spea(ing
A;>
-
7/26/2019 Seminar M.tech
36/49
-
7/26/2019 Seminar M.tech
37/49
A Te:t#to#/peech /ynthesis /yste&
-
7/26/2019 Seminar M.tech
38/49
TT/ /yste&
Fundamental omponents
Te:t
*re#processing*rosody Concatenation
%ords
-
7/26/2019 Seminar M.tech
39/49
*e,t Pre-Processing*e,t Pre-Processing
'nut
Str!ng of characters (sentence)
utut Str!ng of d!hone symbo&s
bect!*e
+erform sentence &e*e& ana&ys!s
+unctuat!on mar,s
+auses bet-een -ords
Con*ert a&& !nut to corresond!ng d!hones
-
7/26/2019 Seminar M.tech
40/49
*e,t Pre-Processing 7;loc( $iagram8*e,t Pre-Processing 7;loc( $iagram8
,ord
/eg&enter
Acrony&
Converter
Nu&ber
Converter
,ord to
DiphoneTranslator
(*honeti7ation)
Diphone
Dictionary
.D/
Nu&ber
Converter
-
7/26/2019 Seminar M.tech
41/49
?umber onverter?umber onverter
Re&ace numera&s -!th the!r tetua&
*ers!ons
1// one hundred
0and&e fract!ona& and dec!ma&
numbers
/.2 o!nt t-o f!*e
-
7/26/2019 Seminar M.tech
42/49
&cron"m onverter&cron"m onverter
Re&ace acronyms -!th s!ng&e &etter
comonents
3.B.C. 3 B C
Change abbre*!at!ons to fu&& tetua&
format
Mr. M!ster
-
7/26/2019 Seminar M.tech
43/49
ord Segmenterord Segmenter
D!*!de sentence !nto -ord segments
Sec!a& de&!m!ter to searate segments
(!.e. 4556)Segments can be:
3 s!ng&e -ord
3n acronym 3 numera&
'dent!fy unctuat!on mar,s
-
7/26/2019 Seminar M.tech
44/49
ord *o $iphone onverterord *o $iphone onverter
7Phoneti9ation87Phoneti9ation8+urose
Trans&ate -ords to the!r d!hone
reresentat!onsResource
D!ct!onary of -ords and the!r d!hones
-
7/26/2019 Seminar M.tech
45/49
-
7/26/2019 Seminar M.tech
46/49
Prosod"
Diphone
Retrieval ConcatenationAcousticanipulation
Diphone
Database
.D/
done"es
no
-
7/26/2019 Seminar M.tech
47/49
$iphone #etrieval
$atabase of recorded diphones
.ver" diphone matched %ith t,t file
$istinguished b" t"pe 7'
-
7/26/2019 Seminar M.tech
48/49
onclusion
$iphones ords
Dsing PSO@& at the joining ends .nsures smooth transition
ords Sentence
Stra!ght o!n!ng at the end
o!nts due to resence of
auses
-
7/26/2019 Seminar M.tech
49/49
*&?E D