0% found this document useful (0 votes)
55 viewsDSP Notes 2 PDF
Uploaded by
Carlos Mella-RijoCopyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
55 viewsDSP Notes 2 PDF
Uploaded by
Carlos Mella-RijoCopyright
© © All Rights Reserved
Available Formats
Download as PDF or read online on Scribd
/ 10
[introduction oo B, Clases of DSP Algorithms
A. Relationship of DSP to other Fields. 1. Linear Time Invariant Filtering
(Convo
(2) UR and FR ers (Length oF impulse
response)
(0) Types of filter (esrition of
>) se fequency response)
1. Freguensy selective: LP, HP, BP,
BR
od 2. Wiener and Kalman
4 Inespolaion
5. Filters fr digital communication,
‘Common topes oll hese is: ans
function and impulse response, Fourier
related transforms, convolution theorem.
However, BSP isnot ed to mechanical
systems, hardware and real ime
processing.
2. Feature Extraction Algorithms ‘3. Statistical Sigal Processing (EESSS2-
(EESSST-Statistical Pattern ‘Spring), A Data Seience Course
‘Resogaiton) (a) Parameter timation: Generate large
(a) Investigate appropriate inca and FVsX,
nonlinear tansformations, (6) Derive conventional or neural net
(b) Transform a input data such estimates ofthe parameters. Example
images, into large “Feature Vectors” power spect density Paw) of|
X via convolution, Fourier tasform, andor sigrals (aed in ita design)
and other technigucs (©) Bound estimation ewor.
(Find wsel sabes of X
{G) Process X inte decisions 4. Algorithms fr Speech Processing
() Develop FV X using homomorphi
fr fikeing
() Use X for Speech Cong, for
production, rage, and transmission
of speech, orfor
() Speech Recegnition
Ea 1, Review of Relevant Math Topics
A. Summation Basis (More later)
Sreteietereiet (amber unm
1s 1s tet
<1)
Ya a1203e44sa1s
Yn =starseatit0
Example
Se +6= (4-2) (08)
Example
Soper .
Example
46413
R34 2D
B. Properties OF Polynomials
1. Notation
fix)mar tartan? tase? acta
Example
0,
2. Factoring
The eres 2a can be rel or complex.
ms)
3. Polynomial Mut
a= Sos
muna
cis) = Seton’
Find
(ym meas fntions of mmm.
{@)o(a) asa fasion oF be) an aC) Procedure
(1) Ona given sie of th equ sign, make
sue that each summation is over
Aifferent index
Sam (
{2)Moveal smmaion ses
rere ip ed ee
(
(2
Saone | Soom |
sown")
[Noe that ns and ns are constants and are
ot functions of mo a).
{) Equate equal power terms on both
‘ies ofthe equal sign and solve for one of
the indices onthe eight hand side,
(4) quate the left had side cotficient ©
‘th general righthand sie coticien,
using the esto from sep (3),
40)= LZaenb4y= $F embn-m)
‘Note hat polyncmial multiplication is
cuivalen to dsretetine convolution,
Example
Seow «(eer [soe )
242m, =n, +2n;
RES: SD amb”
nem+2i, m=n-2i
cme Sen-o00
C-mport
) Eaey-¥ xe)
ay Pitonnn sti aes,
6) ¥ smny- F Soma
SEN
Goda
6) E feo = ey 0)
5) Exeam=ym)
‘
xenmi=yim. ifs variable Details
$F cane $ Fimo
So ONGN 2N_ NEN
i
i
Me IMe 1M» 5
¥ sine
= (sin) =n)"
itso
L=sin"00) i
=sin 5, Gin"
(145i 60)")
Osi)
, Signals and Thelr Properties
1. Signals are Funetions of
Hrequene..)
A signal is function that maps an
independent variable (such realtor
integer n) oa doenent variable
Examples include continuous time signals
0) ad y(t), and sere ime signals
fo}; la) andy) oe (07), (AT) and
yer),
Sianals canbe complex as
a) Refx(n)} +m soy}
Tayo
whore is the magnitude of x and arg)
isan idealized phase of as we'll se ater.
Example: Ssinins) =25[(0") 0°)
leone pagan
Te Ie
—
san),
co
Bample: Ser" eee Sioe'y
eye OO ipa a
w= cao
2. Some Siual Properties
(41) Signals canbe discret tie (40) ),
‘continuoustime (3) discrete
‘space (sith msg fia) te.
(2) Signa canbe infinite length so that
a) as vakes fr 0S 2007
(6) Signals canbe Fite length so that
‘aus Vales for Ni m= Ns for
finite Nand Usefl Functions Unit step Pa
Kronecker Det
I=
0) fp csi ;
yn-ay=[tif 0-3-0 .
fans)
we Eaten) poole
=
Give ington that specie x0) ~
xi0)= mam)
Av) 8) +? -5(0)+(0-1) HD) hy
Example ‘Sinusoidal Sequences
Sketch for a < (0) ~ A-cos(w,n-+6)
x(a)
Example
Complex exponential sequence
‘a= Re® Complex exponential
sequence Period Sequences Ex. For each ofthe following signals, state
hater oF nt iis periodic is, give the
(0)= X(0*N) forall. N= period petod
(a) ecosta)
[Navas beam integer. oer
(6) sins sen)
(@) hiayteostcn'stymyrutn)
(6) w(rjeost227)9)
( yayreos(229)0)
Example Let x(n) ~ As couwon * 6)
‘Under wha condition is x0) periodic?
Acos(won +4)
coswdaeN) +4)
WN = 2a
(1) We ~2RKIN given,
2) N= ekg, Pari if isan
Imeger
Ramp Signal Jom 5 Sinusole
a) = nua) Ea x(a) =sn(vom)fulo-n) ula]
Chiep
1
gular Signal
Goal yenerate a signal whose frequency
‘ares om wy aime 19 Ws a time ns
where
Un) = oem )2aoe 9+ 0)
How arom, mands
related? For large m,
‘whore each amp is
tamed on, (oa) = (ea) 50 ‘The instantancons frequency of x(n is
Aten) 24mm (aes) = 0 and 2antb. we want this fequency 1 vary
ay+2ne-as = O and my =2ns =m) from tows, we have
2enib=w
‘Square Pulse mtb =
la) un) - usenet) an BML | gow, 2am,
Tornsmem a fm 2104-0)
0)~ sinan®sbnéeyulrenpulomt))
wo F. Convolution Operation
‘Continuous and Discrete Time Cases
0 = Pate-urta = Fences
sine Sonik) = $ oka)
‘Sketch: Contauous Time Convolution
Pscudocode for Discrete Time
‘Convolution of Causal Signals
(41) Standard Ofine Consolation
y= Soanca-ky =F oak
PrewdCade
Py
ee
m2 0 and m= N,)Then
(0) =y(0) + Hato)
Encit
nd k
Enda
‘Skt Diseret Time Convolution
a
aa
(2) Altenate OFine Convolution
N=NtN
Fords sNs
Forms Ny
alam) = yinin + him)-xa)
End
End
(6) Real-Time Version fo Finite Length
a)
ea: Store several delayed samples of
x(a) na vector X, which an be caled a:
(2) Inermdiate signal vector
() State vector
(6) Feature Vest For0
