Lecture 16

Signals and Systems

(ELL205)
Prepared using the slides of Prof. Abhishek Dixit
Dept. of Electrical Engineering
IIT Delhi
Inner Product Space
and
Direction
• (If and are real)
• Two vectors are orthogonal if
• Cauchy-Schwarz inequality

Equality is satisfied when

Outline
• Introduction to vectors
• Introduction to inner product and projection theorem
• Introduction to signals as vectors
ℒ2 space is a vector space
• ℒ2 space is a vector space
Is ℒ2 space a vector space?
Addition:
Scalar multiplication:

Is also a finite energy signal?

∞ ∞ ∞

∫ |𝑢 ( 𝑡 ) +𝑣 (𝑡 )| 𝑑𝑡 ≤ ∫ 2|𝑢 ( 𝑡 )| 𝑑𝑡 + ∫ 2|𝑣 ( 𝑡 )|
2 2 2
𝑑𝑡 <∞
−∞ −∞ −∞

Is αu also a finite energy signal?

∞ ∞

∫ |α 𝑢 ( 𝑡 )|
2
𝑑𝑡 = ∫ α |𝑢 ( 𝑡 )| 𝑑𝑡 <∞
2 2

−∞ −∞
Minimum error in approximation
% %
x c y e x  c y
x ⃗
𝑒 ⃗
𝑒 %
e
θ
c y  x cos 
90 o

c y
% 2
c y  x y cos 

% 2 ⟨⃗ 𝑦⟩
𝑥 ,⃗
c y  x. y 𝑐= 2
‖ 𝑦‖
⃗
Minimum error in approximation in
real signals
Estimating x in terms of y x(t ) cy (t )
t2
2
Defining error signal e(t )  x(t )  cy (t ) and its energy Ee (c)  ( x(t )  cy (t )) dt
t1
𝑡2
dEe (c)
Differentiating energy 0 to get ∫ 2 ( 𝑥 ( 𝑡 ) −𝑐𝑦 ( 𝑡 ) ) 𝑦 ( 𝑡 ) 𝑑𝑡 =0
dc 𝑡1
t2
x(t ) y (t )dt
t1
From above we calculate c as: c 
t2
2
 (t )dt
y
t1
Equivalence of vectors and real
signals
t2
⟨⃗ 𝑦⟩
𝑥 ,⃗ x(t ) y (t )dt
𝑐= 2
‖⃗
𝑦‖ t1
c
t2
2
 y (t )dt
t1
𝑡2 𝑡2
𝑦 ⟩=∫ 𝑥 ( 𝑡 ) 𝑦 ( 𝑡 ) 𝑑𝑡 𝑦‖ =∫ 𝑦 ( 𝑡 ) 𝑑𝑡
2
⟨⃗
𝑥 ,⃗ ‖⃗ 2

𝑡1 𝑡1
Try it yourself !!

1 𝑥 ( 𝑡 )=𝑐𝑠𝑖𝑛𝑡
0 π 2π
-1 What is c for minimum error?

1) 2)
4 2
𝑐= 𝑐=
π π
3) 4)
1 8
𝑐= 𝑐=
π π
Finding c

𝑦 ( 𝑡 )=𝑠𝑖𝑛𝑡
2π
𝐸 𝑦 =∫ sin 𝑡 𝑑 𝑡 =π
2

[ ]
2π π 2π
1 1 4
𝑐= ∫
𝐸𝑦 0
𝑥 ( 𝑡 ) 𝑠𝑖𝑛 𝑡 𝑑 𝑡 =
π
∫ 𝑠𝑖𝑛𝑡𝑑𝑡 +∫ − 𝑠𝑖𝑛𝑡𝑑𝑡 =
π
0 π

4
𝑥 (𝑡 )≅ 𝑠𝑖𝑛𝑡
π
 Minimum error in approximation in
complex signals
Estimating x in terms of y x(t ) cy (t )
t2
Defining error signal e(t )  x(t )  cy (t ) and its energy Ee (c)  | x(t )  cy (t ) |2 dt
t1
Algebraic workout:
𝑡2
𝐸 𝑒 ( 𝑐 ) =∫ ( 𝑥 ( 𝑡 ) − 𝑐𝑦 (𝑡 ) ) ( 𝑥 ( 𝑡 ) −𝑐 𝑦 ( 𝑡 ) ) 𝑑𝑡
∗ ∗ ∗

𝑡1

𝑡2
𝐸 𝑒 ( 𝑐 ) =∫ ( 𝑥 (𝑡 ) 𝑥 ( 𝑡 ) −𝑐𝑦 ( 𝑡 ) 𝑥 ( 𝑡 ) −𝑐 𝑥 ( 𝑡 ) 𝑦 ( 𝑡 ) +|𝑐| 𝑦 ( 𝑡 ) 𝑦 ( 𝑡 ) ) 𝑑𝑡
∗ ∗ ∗ ∗ 2 ∗

𝑡1
Minimum error in approximation in
complex signals
𝑡2
𝐸 𝑒 ( 𝑐 ) =∫ ( 𝑥 (𝑡 ) 𝑥 ( 𝑡 ) −𝑐𝑦 ( 𝑡 ) 𝑥 ( 𝑡 ) −𝑐 𝑥 ( 𝑡 ) 𝑦 ( 𝑡 ) +|𝑐| 𝑦 ( 𝑡 ) 𝑦 ( 𝑡 ) ) 𝑑𝑡
∗ ∗ ∗ ∗ 2 ∗

𝑡1

| |
𝑡2 𝑡2 2
1
𝐸 𝑒 ( 𝑐 ) =∫|𝑥 ( 𝑡 )| ∫ 𝑥 ( 𝑡 ) 𝑦 ∗ (𝑡 ) 𝑑𝑡
2
𝑑𝑡 + 𝑐 √ 𝐸 𝑦 −
𝑡1 √𝐸 𝑦 𝑡1
Minimum error in approximation in
complex signals
𝑡2
𝐸 𝑒 ( 𝑐 ) =∫ ( 𝑥 (𝑡 ) 𝑥 ( 𝑡 ) −𝑐𝑦 ( 𝑡 ) 𝑥 ( 𝑡 ) −𝑐 𝑥 ( 𝑡 ) 𝑦 ( 𝑡 ) +|𝑐| 𝑦 ( 𝑡 ) 𝑦 ( 𝑡 ) ) 𝑑𝑡
∗ ∗ ∗ ∗ 2 ∗

𝑡1
2 2 2
|𝑎 − 𝑏| =|𝑎| +|𝑏| − 𝑎 𝑏∗ − 𝑎∗ 𝑏

| |
𝑡2 𝑡2 2
1
𝐸 𝑒 ( 𝑐 ) =∫|𝑥 ( 𝑡 )| ∫ 𝑥 ( 𝑡 ) 𝑦 ∗ (𝑡 ) 𝑑𝑡
2
𝑑𝑡 + 𝑐 √ 𝐸 𝑦 −
𝑡1 √𝐸 𝑦 𝑡1
𝑎 𝑏

2
|𝑎| |𝑏|
2

∗ ∗
𝑎𝑏 𝑎 𝑏
 Minimum error in approximation in
complex signals
𝑡2
𝐸 𝑒 ( 𝑐 ) =∫ ( 𝑥 (𝑡 ) 𝑥 ( 𝑡 ) −𝑐𝑦 ( 𝑡 ) 𝑥 ( 𝑡 ) −𝑐 𝑥 ( 𝑡 ) 𝑦 ( 𝑡 ) +|𝑐| 𝑦 ( 𝑡 ) 𝑦 ( 𝑡 ) ) 𝑑𝑡
∗ ∗ ∗ ∗ 2 ∗

𝑡1

| || |
𝑡2 𝑡2 2 𝑡2 2
1 1
𝐸 𝑒 ( 𝑐 ) =∫|𝑥 ( 𝑡 )| 𝑑𝑡 + 𝑐 √ 𝐸 𝑦 − ∫ ∫
2 ∗ ∗
𝑥 ( 𝑡 ) 𝑦 (𝑡 ) 𝑑𝑡 − 𝑥 ( 𝑡 ) 𝑦 (𝑡 ) 𝑑𝑡
𝑡1 √ 𝐸 𝑦 𝑡1 √ 𝐸 𝑦 𝑡1
Minimum error in approximation in
complex signals
| || |
𝑡2 𝑡2 2 𝑡2 2
1 1
𝐸 𝑒 ( 𝑐 ) =∫|𝑥 ( 𝑡 )| 𝑑𝑡 − ∫ ∫
2
𝑥 ( 𝑡 ) 𝑦
∗
( 𝑡 ) 𝑑𝑡 + 𝑐 √ 𝐸 𝑦 − 𝑥 ( 𝑡 ) 𝑦
∗
(𝑡 ) 𝑑𝑡
𝑡1 √ 𝐸 𝑦 𝑡1 √ 𝐸 𝑦 𝑡1

𝑡 2

∫ 𝑥 ( 𝑡 ) 𝑦 ∗ ( 𝑡 ) 𝑑𝑡
𝑐= 𝑡 1𝑡 2
∫| 𝑦 ( 𝑡 )|
2
𝑑𝑡
𝑡 1
Equivalence of vectors and complex
signals
𝑡 2

⟨⃗ 𝑦⟩
𝑥 ,⃗ ∫ 𝑥 ( 𝑡 ) 𝑦 ∗ ( 𝑡 ) 𝑑𝑡
𝑐= 2 𝑐= 𝑡 1𝑡 2
‖⃗
𝑦‖ ∫| 𝑦 ( 𝑡 )| 𝑑𝑡
2

𝑡 1

𝑡2 𝑡2
𝑦 ⟩=∫ 𝑥 ( 𝑡 ) 𝑦 ( 𝑡 ) 𝑑𝑡 𝑦‖ =∫| 𝑦 ( 𝑡 )| 𝑑𝑡
∗ 2 2
⟨⃗
𝑥 ,⃗ ‖⃗
𝑡1 𝑡1
The Inner Product space of ℒ2
waveforms

= (Energy of the signal)

Properties of Inner Product
a) Hermitian symmetry:
Proof:

b) Hermitian bilinearity:
Properties of Inner Product
c) Strict positivity: with equality if and only if
if Axiom is not satisfied !!
Properties of Inner Product
c) Strict positivity: with equality if and only if
if Axiom is not satisfied !!

Vectors in ℒ2 space are not functions but are equivalence class (all
indistinguishable functions belong to same equivalence class)
Outline
• Introduction to vectors
• Introduction to inner product and projection theorem
• Introduction to signals as vectors
• Signal spaces
Signal Orthogonal Vector Space

N
x(t )   cn xn (t )
n 1

t2 0 m n 
 xm (t ) xn (t )dt  
t1  En m n 

If , orthonormal basis set

Error in approximation
N
e(t )  x(t )   cn xn (t )
n 1
For minimum error cn is
t2 N
Ee  | x(t )   cn xn (t ) |2 dt t2
t1 n 1  x(t ) xn (t )dt
t1
cn 
Ee t2
2
0  xn (t ) dt
cn
t1
What can we choose as basis
vectors?
t2 0 m n 
 xm (t ) xn (t )dt  
t1  En m n 

1. Complex Exponentials
2. Walsh functions, e.g., CDMA applications
3. Legendre Polynomials, e.g., spherical geometries
4. Laugerre functions, e.g., data compression
5. Hermite Polynomials, e.g., interpolation
6. Bessel functions, optical fiber communication (cylindrical geometries)
7. Chebyshev Polynomials, e.g., filter designing
8. Jacobi Polynomials, e.g., data compression and filter designing