EE432

Digital Signal Processing

Introduction
Prof. Young-Gyu Yoon
School of EE, KAIST, 2023

1 EE432 lectures notes include the materials adopted from discrete-time signal processing (textbook), lecture notes by Profs. Munchurl Kim and Junil Choi, and Stanford EE264 course materials.
Instructor & TAs

• Instructor: Prof. Young-Gyu Yoon

• Email: [email protected]
• Office: N24 (LG Hall), #4103-1

• Teaching assistants
• Kim, Gyuri (head TA) [email protected]
• Eom, Minho [email protected]
• Kim, Soi [email protected]
• Ahn, Sungjin [email protected]
Instructor: Young-Gyu Yoon

• B.S. KAIST
• M.S. KAIST
• Mixed-signal circuit (analog-to-digital converter design)
• Ph.D. MIT
• Neuro-engineering (optical imaging & computing)

• Research engineer @ KAIST Institute

• Postdoc @ MIT
• Associate professor @ KAIST
Instructor: Young-Gyu Yoon

• Principal investigator of Neuro-Instrumentation and Computational

Analysis (NICA) Lab

Optical
Microscopy

Machine Brain
Learning
Course logistics

• Textbook
• Alan V. Oppenheim and Ronald W. Schafer,
Discrete-Time Signal Processing, 3rd ed., Prentice Hall.

• Pre-requisites
• EE202 Signals and Systems

• Class website
• Signals and Systems EE432(B) @ http://klms.kaist.ac.kr
• Lecture notes will be posted here
Course logistics

• Class hours
• 9:00AM – 10:30PM (Mon/Wed)
• Class room
• Room #201, ITC Building (N1)

• TA office hour
• To be determined & announced
Course logistics

• Exams
• Three hours (both midterm and final)
• Hand-written two-sided cheat sheet (one A4 paper) allowed
• No electronic device allowed

• Grading
• Midterm exam (40%), Final exam (60%)
• Mid-term exam will be about what we learned in class, but we will
assume that you are familiar with the important concepts you
learned from Signals and Systems
• Final exam will be mainly about what we learn after the mid-term
exam, but we will still assume that you are familiar with the
concepts you learned before mid-term
EE432 Q&A

1. Please Make your questions public, not private.

2. Please Classum @ KLMS (no e-mail or Q&A board)

• Never use Q&A board @ KLMS

• Use email only for the private matters (e.g., about grading)
EE432 Q&A rules

1. Please follow simple form for Title.

• [Lecture note number, quiz number, etc] short summary

• [Lecture02] bilateral vs. unilateral z-transform

2. Explain your question well so that we can help you efficiently.

3. Check others’ question. You can get answer or hint. Please do not make duplicate
question.
 전기 및 전자공학부 학업 윤리 규정 제정 시행

- 올바른 학업 윤리의식 함양 및 건전한 면학 분위기 확립

목 - 윤리위반행위에 대한 경각심 고취를 통한 윤리위반행위 사전 예방
적 - 윤리위반행위 발생 시 체계적인 대응 및 재발방지
- 각 수업에서 허용하는 수업자료의 공유 및 참고 정책 사전 공지
- 정책 위반 행위 발생 시 규정에 명시된 절차에 따라 EE 학생 윤리 위원회에서
적
관련 조사/ 심의/ 학생지도/ 징계 진행
용
※ EE 홈페이지 공지사항에 학업 윤리 규정 전문 게시 https://ee.kaist.ac.kr/node/15358
학업 윤리 위반 행위 관련 신고/문의처 [email protected]
수업자료 공유정책 비고*
가능여부 (o/x)
(현 학기 중 공유/이후 학기 공유를 모두 포함) (수업 별 특수 정책이 있는 경우 명시)
출제된 과제 및 과제풀이의 공유 및 배포

제공된 강의 자료의 공유 및 배포
정책공 출제된 시험문제의 공유 및 배포
지
비고*
이전 수업자료 참고정책 가능여부 (o/x)
(수업 별 특수 정책이 있는 경우 명시)
과제 수행 시 이전 기출 과제 및 과제풀이 참조
과제 수행 시 수강생 간의 토론/협업
시험 준비 시, 이전 기출문제 자료 참조

1
0
학업윤리 위반 사례 공유
위반행위 처분내용 발생 횟수 비고

학부 내 징계
과제물 표절 14
사회봉사 30시간

허용되지 않은 자료 학부 내 징계
9
제공 및 전달 사회봉사 15시간

부정행위 방조 및
경고 (경고메일 발송) 3
허용되지 않은 자료 단순 열람

학부 내 징계 튜터링 과정에서 튜티에게 부정행위를 권유

튜터의 윤리위반 2
사회봉사 30시간, 및 지원한 경우

전기및전자공학부 과목에서 발생한 부정행

유기정학 2개월,
2 위
사회봉사활동 100시간
EE학생 윤리규정에 따라 처분.
시험부정행위
학부 내 징계 타과 과목에서 발생한 부정행위, 과목 담당
1
사회봉사 60시간 교수가 학생 징계를 요청하지 않은 케이스

진상조사/면담/심의를 진행하였으나,
시험부정행위
- 1 제보내용 외 부정행위를 확정할 수 있는 근
(익명제보)
거가 없어 처분하지 않음.
Policies
Sharing distributed material (present semester and after)
Permission Note*
Exam solutions O* *only private sharing allowed (i.e., do not
upload on internet)
Lecture material O* *only private sharing allowed (i.e., do not
upload on internet)
Exams questions O* *only private sharing allowed (i.e., do not
upload on internet)

Reference material
Permission Note*
Previous assignments O
& solutions
Discussion among O* *Students are encouraged to
students discuss
Previous Exams O

https://ee.kaist.ac.kr/en/node/15401
 Digital processing of analog signals

xc(t) x[n] Digital Signal y[n] yc(t)

ADC DAC
Processor

• Analog-to-digital converter (ADC)

• Performs filtering, sampling, and quantization
• Digital signal processor
• Performs some operation e.g., filtering, FFT, etc
• Digital-to-analog converter (DAC)
• Performs quantization and reconstruction (filtering)

9/44
 Discrete-time signals

xc(t)
3 x[n] = xc(nT )

T 2T 3T . . . t
x[n] is only defined for n ∈ Z

• Discrete-time signals (or simply sequences) may be inherently discrete or

they may be obtained by sampling a continuous-time signal.
11/4
4
 Unit impulse (Kronecker delta) function

• The role similar to the unit impulse function in continuous-time

signals

15
 Important role of unit impulse

• An arbitrary sequence can be represented as a summation of

scaled-and-shifted unit impulse functions

16
 Unit step function

• Another interpretation of unit step sequence

17
 Exponential sequences

positive A

• General form:

• What will x[n] look like with ?

18
 Sinusoidal sequences

• With

frequency phase

19
 Properties of exponential and sinusoidal sequences

• Because n is always integer, with arbitrary integer r,

period
• Periodicity

• For exponential sequence to be periodic, i.e., for

all n, it is necessary and sufficient to have

may not be periodic depending on frequency

20
 Periodicity examples

• is periodic with period N=8 different from

continuous-time
sinusoidal signals
• is periodic with period N=16

• is not periodic at all

21
 Number of distinguishable frequencies

• Let

same frequency in exponential sequences

• There are N distinguishable frequencies that are periodic with

period N
• Number of basis functions needed for discrete-time Fourier analysis

22
 Classification of discrete-time signals

Energy signals have finite energy

Power signals have infinite energy (e.g., periodic signals), but they have finite average
power

15/4
4
 Symmetry properties of sequences

(even symmetry)
(odd symmetry)

Any signal x[n] can be decomposed as a sum of an even (xe[n]) and an odd (xo[n]) component:

where
(even component)

(odd component)

16/4
4
 Discrete-time systems

Discrete-Time
System

Linearity or superposition

Time invariance or shift invariance

A time shift of the input causes an equal time shift of the output

17/4
4
Linear time-invariant (LTI) systems
δ[n] h[n]

1 1

n n

LTI System
δ[n] h[n]

LTI systems are completely characterized by their impulse response.

18/44
Linear time-invariant (LTI) systems
δ[n − 1] h[n − 1]

1 1

n n

LTI System
δ[n − 1] h[n − 1]
from time invariance

18/44
Linear time-invariant (LTI) systems

1 1

n n

LTI System
δ[n] − 0.5δ[n − 1] h[n] − 0.5h[n − 1]
from linearity

18/44
Linear time-invariant (LTI) systems

1 1

n n

LTI System
δ[n] − 0.5δ[n − 1] h[n] − 0.5h[n − 1]

The convolution

Shorthand notation: y[n] = x[n] ∗h[n] or y[n] = (x ∗h)[n]

18/44
 Linear time-invariant (LTI) systems

• LTI systems have significant signal-processing applications

• Recall that the input sequence can be represented as

• Let
• Output becomes

Linear

Time-invariant

30
 Properties of convolution sum

• Commutative

• Distributive

• Associative

• Combining commutative and associative

31
Properties of convolution: commutative

Proof.
Properties of LTI system: commutative

Interpretation: the roles of input and impulse response are interchangeable

Properties of convolution: distributive

Proof.
Properties of LTI system: distributive

+
Properties of convolution: associative

Proof.

Interpretation: order of convolution operation does not matter

Properties of LTI system: associative
Equivalent LTI systems

associative

commutative

associative
 Causality

• For every choice of , the output sequence value at the index

depends only on the input sequence values for

• Current output is a function of only past and present inputs, not future
inputs

• Examples
• Backward difference system is causal

• Forward difference system is non-causal

39
 Stability

• Many different definitions of ‘stability’ exist

• We focus on bounded-input bounded-output (BIBO) stability

• A system is BIBO stable iff every bounded input sequence produces

a bounded output sequence.

• Input is bounded if there exists a fixed positive finite value such

that

With bounded input, there exists a fixed positive finite values such
that

40
 Stability examples

• Squaring system is BIBO stable

• Accumulator is NOT BIBO stable

Proof: Check with the input

41
 Important properties of systems

• Memoryless systems

• Linear systems
Two properties define LTI systems
What about other properties?
• Time-invariant systems

• Causality

• Stability

42
 BIBO stable and causal LTI systems

• LTI systems are stable iff

• LTI systems are causal iff

• LTI systems are in general with memory

43
 Impulse responses of basic LTI systems
• Ideal delay

• Moving average

BIBO stable?
Causal?
• Accumulator

• Forward difference
• Backward difference

44
 IIR vs. FIR systems

• IIR: infinite-duration impulse response

• IIR examples

• FIR: finite-duration impulse response

• FIR examples
• Ideal delay, moving average, forward/backward differences
• Always stable in general (why?)

45
 Closer look into ideal delay systems
• The output of delay system

• The impulse response of delay system

• Therefore,

Convolution of a shifted impulse sequence with any signal x[n]

is easily evaluated by simply shifting x[n]

• Example:

46
 Concept of inverse systems

• Note

• The system is inverse of if

• Example

47
 Linear constant-coefficient difference equations

• Some LTI systems can be represented using linear constant-

coefficient difference equations

• Rearrange to have a recursive form

Past outputs Present and past inputs

= feedback path

48
 Difference equations example

• Accumulator

49
 Difference equations example

• Moving average with

• Alternatively representation

50