أنظمة التعرف على الصوت - speech recognition systems
TRANSCRIPT
-
8/14/2019 Speech recognition systems -
1/16
1
:
. .
info.kutub.www
-
8/14/2019 Speech recognition systems -
2/16
2
,
... ,
.....
XP
.....
:
(Computer Science -1 CS)
(Information System -2 IS)
: ((CS
( (Image Processing
( (Speech Processing
((Speech Processing
) ( ((speech recognition
: ((Speech
: ((voice.
1000
.
.
-
8/14/2019 Speech recognition systems -
3/16
3
( )
.
((speech recognition .
((speech recognition system :
3 :
1- Pre-processing
2- Recognition
3-Communication
Pre-processing
((Recognizer
(( s/w & h/w ((Recognizer
h/w s/w
. Pre-processing
Analog signal
-: Continuous Signal
-
8/14/2019 Speech recognition systems -
4/16
4
.
discrete signal :
binarydigital signal discrete signal form
Quantization
((threshold ((2level :
(Discrete Signal ((Analog Signal )
((Quantization ( (Processing
.
....
, Amplitude : amplitude amplitude .....
....
-
8/14/2019 Speech recognition systems -
5/16
5
level threshold * threshold
.... thresholdimage threshold
(Speech Recognition System )
: pre-processing - recognition - communication
((analog signal pre-processing 0 , 1 (digital signal) ...............
recognition:
:
(identification & verification)
verificationidentification
: identification
,x x x
x
x Distance measurement x
.. x
Distance measurement identification
x .. x verification
verification x ( x
recognition Communication
-
8/14/2019 Speech recognition systems -
6/16
6
Communication:
H/WS/W :
Security ,
" " " "
!! security
white noise
. noise
:: noise
education speech ..
control ..
Diagnosis ..
Bio informatics
speech processing ,,,
.... image processing
pre-processing , recognition , communication :
-
8/14/2019 Speech recognition systems -
7/16
7
..
..
.pre-processing :
((Data collection & acquisition /1
..
((voiced & unvoiced detection 2/
...
amplitude ),,(
),(
noise
Unvoiced Voiced Zero Crossing
-
8/14/2019 Speech recognition systems -
8/16
8
voiced
amplitude
" horizontal access "
zero crossing , zero
voiced speech Zero Crossing
unvoiced
, amplitude
...... zero crossing
end -point-detection /3
Processing noisesignal
noise filter noise ...
4/WrappingTime
( ) :
............ algorithm
-
8/14/2019 Speech recognition systems -
9/16
9
Framming /5
. 20
, ) )20 speech )20 frame frames speech
frame ((sampleframe ) frame
.processing
frames .... ..........
Windwing/6 .....frames
frame frame window
50% frame 50% 50%windows --:
-
8/14/2019 Speech recognition systems -
10/16
10
7/Modeling
Analog Signal Speech Signal features modeling .....
Feature extraction 8/
modeling ...
..........
:
isolated word recognition I W R co-articulation
..
connected word recognition C W R .. Stops
continuous speech recognition C S R
..
Speech understanding S U ..
speaker identification ,speaker verification S I, S V .
word spotting w s key word
-
8/14/2019 Speech recognition systems -
11/16
11
...................
neural signal Waveform gestures
.. .utterance
Waveform(speech).., Waveform
Energy time Waveform.Amplitude
spectrum Amplitude frequencies :
. processing
time:3 Spectrogram frequencyamplitude wave.
utterance
vocal folds or vocal cords vocaltract
oral cavity
-
-
8/14/2019 Speech recognition systems -
12/16
12
nasal cavity .Acoustic wave form
Waveform acoustic
.
.
3 vocal apparatus Throat mouthnose
. vocal cords throat
:
velumTongueteeth: Hard palate: roof of the mouth alveolar
lips
.
periodic voiced
.
-
8/14/2019 Speech recognition systems -
13/16
13
quasi periodic voiced
like periodic: noise
randomnoise
:
.
:
3 .
voiced sound 1
frequency
:
-
8/14/2019 Speech recognition systems -
14/16
14
: frequency
:
.. :
.. :
400 - 60 .100 180
unvoiced sound 2
voiceless sound noise
.
3 ....
:
Glottis ....
:
phoneme .
-
8/14/2019 Speech recognition systems -
15/16
15
linguistics ... phoneme
.....
phonetics ...
voiced
.unvoiced
.
consonantvowel
vowela, e ,u , o , aa ,ee , au ,uh
consonant .
consonant
place of articulation 1
..
articulationmanner of 2
:
-
8/14/2019 Speech recognition systems -
16/16
16
plosive G bid B
gate :
Fricatives
:
v , th , z, f, s, sh
Nasals m, n, ng nasal cavity.
Affricates Judge J
Approximant :
semivowels W, Y
Liquids R, L