Chapter-2 1

Chapter-2

Fourier series and Geometry
2.0 Introduction
Deemed one of the crowning achievements of the 20th Century, the Fourier Series has
applications that are far reaching in various fields of science and mathematics. The
Discrete Fourier Transform is one particular tool widely used in todays age of computers
and solid state electronics. From graphic equalizers in stereos to the most advanced
scientific sampling software, the usefulness of this mathematical feat is astounding. Most
of the readers of this book might already know this fact but many of them may not be
knowing that Fourier series as a mathematical tool in analyzing signals has strong
connection with geometry. In this chapter our aim is to understand the theory of Fourier
series from a geometrical view point.

We assume that you are familiar with vector algebra and co-ordinate geometry. If you are
really comfortable in using these topics, you can very well understand what Fourier
Series is, computation and interpretation of Fourier Series coefficients, Fourier
Transform, Discrete Fourier Transform, Fast Fourier Transform etc. There exists a
strong analogy between what you do in vector algebra and what you do in signal
processing. This analogy will help you to visualize and give interpretation to the
processes and output of the processes that you do on signals. Development of this
geometrical mental picture about signal processing is the main aim of this chapter.
Einstein once said Imagination is more important than knowledge. Signal processing ,
which many students consider as dry and non-intuitive, demands imagination and some
form of abstract thinking from the part of students. In fact, it requires only that. Once you
develop a conceptual link between geometry (vector algebra) and signal processing, you
need not remember any formula and every formula that you come across will become
transparent to your mind. Here we will refresh concepts from vector space. The detailed
exposition of vector space is given in Appendix-A. In this introductory chapter, we
wont try to be 100% mathematically precise in our statements regarding vector space.
Instead our aim is to relate geometry and Fourier Series in an intuitive way.
Chapter-2 2
2.1 Vector space
Any vector in 3-dimensional space can be represented as k c j b i a V

+ + = . k j i

, ,
are unit vectors in three orthogonal directions. This orthogonality is expressed by
requiring the condition that 0 . . . = = = k i j k j i

. The orthogonality condition ensures
that the representation of any vector using k j i

, , is unique.
2.1.1 Bases. We call k j i

, , as bases of space
3
9 . By this we mean that any vector in
3
9 can be represented using k j i

, , vectors. We also say k j i

, , span the space
3
9 . Let
k c j b i a V

+ + = be a vector in
3
9 . Continuously changing scalar a,b,c; we will
go on get new vectors in
3
9 . We imagine that, set of all such vectors constitute the
vector space
3
9 . We express this truth by

3
, ,
) { 9 + + k c j b i a Span
c b a


2.1.2 Orthonormality. Norm of a vector k c j b i a V

+ + = is conventionally
defined as
2 2 2
c b a + + and denoted as V . V is equal to V V

- . We interpret this
quantity as the magnitude of the vector. It must be noted that other definitions are also
possible for norm. What is interest to us is that our basis vectors of
3
9 , i.e., k j i

, , are
having unity norm. Hence the name unit vectors. They are also orthogonal. Therefore
they are called orthonormal vectors or orthonormal bases. How does it help us?. Given a
vector k c j b i a V

+ + = , we can easily tell to which direction k j i

, , , the vector
V

, is more inclined by noting the value of a,b,c.

2.1.3 Projection. Given a vector V

how shall we find its normal component vectors, or

in other words how shall we find the scalar coefficients a, b and c ?.. We project V

on to
bases k j i

, , to get a, b and c. This is the direct result of orthogonality of our basis
vectors. Since k c j b i a V

+ + = ,
a i k c j b i a i V = - + + = -

) (
b j k c j b i a j V = - + + = -

) ( and
c k k c j b i a k V = - + + = -

) (
Chapter-2 3
So projecting vector V

on to a base of space
3
9 (spanned by a set orthonormal vectors)
gives the corresponding coefficient (component) of the base. There is a parallel for this
in Fourier Series analysis.
With these ideas firmly engraved in your mind you are ready to understand and visualize
Fourier series.
2.2 Functions and Function spaces.
In vector space, we represent a vector in terms of orthogonal unit vectors called bases.
Fourier series is an extension of this idea for functions (electronics engineers call them as
signals). A function is expressed in terms of a set of orthogonal functions. To correlate
with ideas in geometry, we now need to introduce the concept of orthogonality and norm
with respect to functions. Note that, function is quantity which varies with respect to one
or more running parameter, usually time and space. Orthogonality of functions depends
on multiplication and integration of functions. Multiplication and integration of function
must be thought of as equivalent to projection of a vector on to another. Here comes the
requirement for abstract thinking, mentioned at the start of the chapter.

2.2.1 Orthogonal functions:- Two real functions ) (
1
t f and ) (
2
t f are said to be
orthogonal if and only if dt t f t f ) ( ) (
2 1
}

=0 (2.1)
Ponder over this equation. What we are doing?. We are first doing a point by point
multiplication of two function. Then we are summing up the area under the resulting
function (obtained after multiplication). If this area is zero, we say, the two functions are
orthogonal. We also interpret it as similarity of functions ) (
1
t f and ) (
2
t f . If ) (
1
t f and
) (
2
t f are vary synchronously, that is if ) (
1
t f goes up ) (
2
t f also goes up and if ) (
1
t f
goes down ) (
2
t f also goes down then we say the two functions have high similarity .
Otherwise, they are dissimilar. Magnitude or absolute value of dt t f t f ) ( ) (
2 1
}

will be
high if they vary synchronously and zero if they vary perfectly asynchronous away.
Chapter-2 4

Figure 2.1 orthogonal functions
For example consider two functions ) (
1
t f =sin(t) and ) (
2
t f =cos(t) for t in the interval
(0,2t). Figure 2.1 shows the two functions and their point wise product. Observe that the
net area under the curve sin(t)cos(t) is zero. That is, the area above the t axis is same as
the area below the t axis. For these two functions, it can be easily shown that, for t in the
range of multiples of 2t , area under the product curve is always zero. Therefore , we say
that sin(t) and cos(t) are orthogonal in the range (0,2t). Specification of range is very
important when you say orthogonality of functions. Because the deciding factor is area
under the curve which in turn depends on the range of integration. Now here observe the
variation of sin(t) and cos(t). When sin(t) is increasing, cos(t) is decreasing and vice
versa. For range of t in multiples of 2t, the two functions considered are said to be
orthogonal.
The geometrical analog of dt t f t f
t
) ( ) (
2 1
}
=0 is the dot product, 0 . = j i

. For the given
interval, when ) (
1
t f is projected on to ) (
2
t f (projection here means multiplying and
integrating), if the result is zero , we say the two functions are orthogonal in the given
interval. Clearly look at the analogy. Taking a dot product in vector space(multiplying
corresponding terms and adding) is equivalent to multiplying and integrating in the
function space (point wise multiplication and integration). The dot product is maximum
Chapter-2 5
when two vectors are collinear and zero when they are at 90 degree. This analogy will
help you to easily visualize and interpret various results in signal processing.
2.2.2 Orthonormal Functions. Two real functions ) (
1
t f and ) (
2
t f are said to be
orthonormal if and only if dt t f t f ) ( ) (
2 1
}

=0 and 1 ) ( ) ( =
}

dt t f t f
i i
, for i=1,2
2.2.3 Function Spaces. Just like unit orthogonal set of vectors span vector spaces,
orthogonal set of functions can span function spaces. Consider following function defined
over a period T.
Let T / 2t e = (2.2)
1) u(t)=1 for T t s s 0
2) ) . cos( t ne , ,.....; 2 , 1 = n T t s s 0
3) ) . sin( t ne , ,.....; 2 , 1 = n T t s s 0











Figure 2.2 Orthogonality of functions

Here we have infinite number of functions of sines and cosines. These set of functions
are mutually orthogonal in the interval (0,T). Let us verify the truth through some
representative examples.
Let us take functions u(t) =1 and ) . sin( t e for T t s s 0 . These functions and their
product function is shown in figure 2.2. Point wise multiplication gives the resulting
Chapter-2 6
curve as ) . sin( t e . What is the area under the curve ) . sin( t e in the interval (0,T)?.
obviously zero. Mathematically,
0 ) / 2 sin( . 1 ) . sin( ) (
0
= =
} }
= =
dt T t dt t t u
T
t
T
o t
t e
We say u(t)=1 is orthogonal to ) . sin( t e in (0,T). Do not forget the relation between e
and T. Similarly we can show that u(t)=1 is orthogonal to ) . cos( t e in (0,T).
In general,
}
=
=
T
t
dt t n t u
0
0 ) . sin( ) ( e also
}
=
=
T
t
dt t n t u
0
0 ) . cos( ) ( e for all n=1,2,.
Consider now ) . cos( t e and ) . sin( t e in (0,T). Figure below shows the two functions and
corresponding function obtained after point wise multiplication. Area above and below of
the resulting curve is same implying that net area is zero. This in turn imply that
) . cos( t e and ) . sin( t e in (0,T) are orthogonal. Figure 2.3 illustrates the concept.

Figure 2.3. plots of sin(wt), cos(wt) and sin(wt)cos(wt)
Chapter-2 7
Generalizing, 0 ) . cos( ) . sin(
0
=
}
=
T
t
dt t m t n e e , for all m,n =1,2,3,
Similarly 0 ) . sin( ) . sin(
0
=
}
=
T
t
dt t m t n e e for all integer m and n such that n m =
Also 0 ) . cos( ) . cos(
0
=
}
=
T
t
dt t m t n e e for all integer m and n such that n m =
Thus we have a set of functions defined over (0,T) such that they are mutually
orthogonal.
2.2.4 Orthogonal basis functions
In vector algebra , we know that the unit orthogonal vectors k j i

, , span the space
3
9 .
This imply that any vector in
3
9 can be represented by unit vectors or bases k j i

, , .

Similarly the set of functions,
1) u(t)=1 for T t s s 0
2) ) . cos( t ne , ,.....; 2 , 1 = n T t s s 0
3) ) . sin( t ne , ,.....; 2 , 1 = n T t s s 0
defined over T t s s 0 span a function space and we denote it as
2
L . Most of the
practical signals or functions are assumed to belong in the space of
2
L . This means that,
practically any continuous signal in the interval (0,T) can be represented using the above
bases. Or mathematically,
. .......... ) . 2 cos( ) . cos( ) ( ) (
2 1 0
2
+ + + = e t a t a a t f L t f e e

+ .......... ) . 2 sin( ) . sin(
2 1
+ + t b t b e e (2.3)

Let us look at the concept from another angle.
Consider the sum ) (t f for each t in the range (0,T) given in (2.3)
Choose a set of coefficients ,.... ,
1 0
a a and ,... ,
2 1
b b and make the function(or signal)
) (t f in the range (0,T) . Now go on choose different set of coefficients to get a different
signal. The set of all such possible signal generated is our signal space
2
L in the range
Chapter-2 8
(0,T). Practically all smoothly varying functions that you can imagine in the interval (0,T)
belongs to
2
L . Or in other words, practically all smoothly varying functions that you can
imagine in the interval (0,T) can be represented by the equation
.......... ) . 2 sin( ) . sin( . .......... ) . 2 cos( ) . cos( ) (
2 1 2 1 0
+ + + + + + = t b t b t a t a a t f e e e e

How many coefficients are required to represent a signal depends on how smooth and
how fast your signal is varying in the range (0,T).
2.2.5 Othonormality and the method of finding the coefficients:- The problem that we
are addressing is finding the coefficient set etc b a b a a .... , , , ,
2 2 1 1 0
. In vector algebra,
orthonormality of basis vectors allowed us to find the component of a vector along a
particular base by projecting the vector on to that base. The property that
k j i = = =1 where i i i

- = , and k j i

, , are orthogonal implies that for a vector V

3
9 e , if , , , c k V b j V a i V = - = - = -

then . k c j b i a V

+ + = It is a unique
representation using the bases k j i

, , . To reiterate the principle, to find a particular
coefficient , project the vector on to the corresponding basis. The same mental picture is
applicable to Fourier series representation of signals.

In the last section, we have found out a set of functions that are orthogonal among
themselves in a given range (0,T). These functions are orthogonal but we havent checked
whether they form an orthonormal set. In function space , orthonormality of two
functions in (0,T) requires that
}
=
T
dt t f t f
0
2 1
0 ) ( ) ( ,
}
=
T
dt t f t f
0
1 1
1 ) ( ) ( and
}
=
T
dt t f t f
0
2 2
1 ) ( ) (

Consider the basis function ) . sin( t ne .

2
)
2
sin( )
2
sin( ) . sin( ) . sin(
0 0
T
dt
T
nt
T
nt
dt t n t n
T T
} }
= =
t t
e e . The result is a function of T and is
not equal to 1. For any integer n>0, the integral is
2
T
. So, we say our sine bases are not
normalized.
Chapter-2 9
Similarly it can be easily shown that,

2
)
2
cos( )
2
cos( ) . cos( ) . cos(
0 0
T
dt
T
nt
T
nt
dt t n t n
T T
} }
= =
t t
e e for n =1,2,. So our cosine bases
are also not normalized bases.
What about our first basis function (which signal processing people call as DC base))
u(t)=1, in (0,T) ?.
The integral T dt
T
=
}
0
2
1 . Therefore, none of our bases are normalized.
So how shall we normalize our bases?. Multiply all the sine and cosine bases with
T
2

and u(t)=1 with
T
1

So that 1
2
.
2
)
2
sin( )
2
sin(
2
) . sin( / 2 ) . sin( / 2
0 0
= = =
} }
T
T
dt
T
nt
T
nt
T
dt t n T t n T
T T
t t
e e
1
2
.
2
)
2
cos( )
2
cos(
2
) . cos( / 2 ) . cos( / 2
0 0
= = =
} }
T
T
dt
T
nt
T
nt
T
dt t n T t n T
T T
t t
e e
1 .
1
. 1
1
/ 1 . / 1
0 0
= = =
} }
T
T
dt
T
dt T T
T T

The set of normalized bases of Fourier series are.

)
`

),... . sin(
2
), . cos(
2
).,...., . 2 sin(
2
), . 2 cos(
2
), . sin(
2
), . cos(
2
,
1
t n
T
t n
T
t
T
t
T
t
T
t
T T
e e e e e e
All these functions are defined in the interval (0,T) and T / 2t e =
Let us now take a function
2
) ( L t f e defined in the interval (0,T) and express it using the
normalized sine and cosine bases defined over the same period (0,T). Note that the length
of bases are same as the length of the signal. Also note that fundamental frequency of the
sinusoidal (also cosinusoidal) series is given by T / 2t e = . (This is some thing which
students tend to ignore or forget and hence the repetition of the fact in many places).
T
a t f
1
) (
0
' = + ) . cos(
2
1
t
T
a e ' + ) . sin(
2
1
t
T
b e ' +..,
Chapter-2 10
+ ) . cos(
2
t n
T
a
n
e ' + ) . sin(
2
t n
T
b
n
e ' + (2.4)
Here ,.... , ......, , ,
2 1 2 1 0
b b a a a ' ' ' ' ' are called normalized coefficients.
Now, given that, such a representation is possible, how shall we find out the normalized
coefficients. The answer is, project ) (t f on to corresponding bases. Suppose we want to
find
n
a' . Then project ) (t f on to ) . cos(
2
t n
T
e
That is
n
a' =
}
T
t n
T
t f
0
) . cos(
2
) ( e . Note the analogy. In vector algebra, to find a component
coefficient, we project the vector on to the corresponding unit vector (base).
Let us verify the truth through mathematical microscope. Multiplying both sides of
equation (1) by ) . cos( / 2 t n T e and integrating over (0,T), we obtain
dt t n T
T
a dt t n T t f
T T
) . cos( / 2 .
1
) . cos( / 2 ) (
.
0
0
0
e e
} }
' =
+ dt t n T t T a
T
) . cos( / 2 ) . cos( / 2
0
1
e e
}
' +
dt t n T t T b
T
) . cos( / 2 ) . sin( / 2
0
1
e e
}
' +..,
+ dt t n T t n T a
T
n
) . cos( / 2 ) . cos( / 2
0
e e
}
' +

dt t n T t n T b
T
n
) . sin( / 2 ) . sin( / 2
0
e e
}
' +
=
n
a'
n
T
a dt t n T t f ' =
}
) . cos( / 2 ) (
0
e since 1 ) . cos( / 2 ) . cos( / 2
0
=
}
dt t n T t n T
T
e e and all
other integral vanishes because of orthoganility property of the bases.
Similarly
n
b' = dt t n T t f
T
) . sin( / 2 ) (
0
e
}
. That is, project ) (t f on to the corresponding
normalized base.
Chapter-2 11
What about
'
0
a ?. Project ) (t f on to the base T / 1 defined over (0,T) , giving
0
a' = dt T t f
T
/ 1 ) (
0
}

Well, we understood how to get normalized coefficients in fourier representation of
signals. To get a particular coefficient simply project ) (t f on to the corresponding
normalized base. Projection in function space means point wise multiplication and
integration of the functions concerned over the defined period.
Once we obtain the normalized coefficients we substitute in equation (2.4) to get fourier
series representation of the signal. Let us simplify the equation (2.4) to avoid scalars
T / 2 and T / 1
Let
0 0
1
a
T
a = ' so that
} }
= = ' =
T T
dt t f
T
dt t f T T a T a
0 0
0 0
) (
1
) ( / 1 / 1 / 1
Note that we substituted the expression for
0
a' to get
0
a
Similarly let
n n
a T a = ' / 2 . Then
} }
= = ' =
T T
n n
dt t n t f
T
dt t n T t f T a T a
0 0
) . cos( ) (
2
) . cos( / 2 ) ( / 2 / 2 e e
Let
n n
b T b = ' / 2 . Then
} }
= = ' =
T T
n n
dt t n t f
T
dt t n T t f T b T b
0 0
) . sin( ) (
2
) . sin( / 2 ) ( / 2 / 2 e e

with this simplification we can rewrite equation (1) as
0
) ( a t f = + ) . cos(
1
t a e + ) . sin(
1
t b e +.., + ) . cos( t n a
n
e + ) . sin( t n b
n
e + (2.5)
where
}
=
T
dt t f
T
a
0
0
) (
1
(2.6)
}
=
T
n
dt t n t f
T
a
0
) . cos( ) (
2
e , n=12,3, (2.7)
}
=
T
n
dt t n t f
T
b
0
) . sin( ) (
2
e , n=1,2,3 (2.8)
Chapter-2 12
These are the formulae given in most text books.

2.2.6 Complex Fourier series

Now we move on to complex bases.
We know that
j
e e
t
t j t j
2
) . sin(
. e e
e

= and
2
) . cos(
.t j t j
e e
t
e e
e

+
=
Now, + ) cos(
1
t a e ) . sin(
1
t b e =
t j
e
jb a
. 1 1
2
) (
e

+
t j
e
jb a
. 1 1
2
) (
e
+

Let us denote )
2
(
1 1
1
jb a
C

= and )
2
(
1 1
1
jb a
C
+
=

The Fourier series can now be rewritten as
...... ...... ) (
. . .
1
.
1 0
+ + + + + + =

t jn
n
t jn
n
t j t j
e C e C e C e C C t f
e e e e

=
t jn
n
n
e C
. e

=
(2.9)
Note that formulation involves complex exponentials and the concept of negative angular
frequency, e . There is nothing imaginary in any signal and there is no negative
frequency. Complex exponentials facilitate easy manipulation of multiplication and
integration involved in the calculation of Fourier series coefficients than multiplication
and integration using plain sine and cosine functions .
Let us first try to visualize
t j
e
. e
, where , / 2 T t e = and T is the period of our signal .
By Demovies formula,
t j
e
. e
= ) . sin( ) . cos( t j t e e + . We can think of this function as a
bundle of sine and cosine function kept together side by side such that they do not mix
with each other. The function of j can be thought of as a separator indicating that the two
parts (viz ) . cos( t e and ) . sin( t e ) are orthogonal .

2.2.7 Orthogonality of complex exponential bases.
The set of bases { } ,......... , , ,......... , , 1
. . t jn t jn t j t j
e e e e
e e e e
are orthogonal over the
interval [0, T], where T / 2t e = , and T is the duration of the signal which we are
representing using the bases. However there is a twist with respect to interpretation of
Chapter-2 13
orthogonality and projection of a function onto a complex function. To make you
prepared for that twist and to accept that twist, let us consider the following.

Let k c j b i a V

+ + = be an element of
3
9 . The norm of this vector ( or magnitude) is
obtained by projecting V

on to itself and then taking the square root of resulting quantity.

That is
2 2 2
c b a V V V + + = =

. But in case of a complex number jb a Z + = , how
will we find the magnitude. Here we have to multiply Z with its complex conjugate Z
and then take the square root. In case of real function, say ) (t f , we find norm by
}
t
dt t f t f ) ( ) ( . In case our function is complex , say ) (t Z , its norm is given by
}
t
dt t Z t Z ) ( ) ( . This has implication in the formula for projection of a function on to a
complex function. We must multiply the first function with complex conjugate of the
second function and then integrate. With this twist in mind, we will prove that the set of
functions { } ,......... , , ,......... , , 1
. . t jn t jn t j t j
e e e e
e e e e
are orthogonal.
We denote the projection of ) (
1
t f on to ) (
2
t f as ) ( ), (
2 1
t f t f
Case 1: u(t)=1 and
t j
e
. e
, for T t s s 0 and T / 2t e =
t j
e
.
, 1
e
= 0 1
0
2
0 0
= = =
} } }

dt e dt e dt e
T
T
t
j
T
t j
T
t
t j
t
e e

case 2:
t jn
e
e
and
t jm
e
e
where n and m are positive integers.
t jm t jn
e e
e e
, = dt e e
T
t
t jm t jn
}
=

0
e e
= 0
0
2
) (
0
) (
} }
= =
+
+
T
t
T
m n j
T
t m n j
dt e dt e
t
e

case 3:
t jn
e
e
and
t jm
e
e
where n and m are either positive or negative integers with n m =
t jm t jn
e e
e e
, = dt e e
T
t
t jm t jn
}
=

0
e e
= 0
0
2
) (
0
) (
} }
= =


T
t
T
m n j
T
t m n j
dt e dt e
t
e

Cases 1, 2, and 3 proves that the functions are orthogonal.
What about orthonormality of complex exponential functions?.
Since,
t jn t jn
e e
e e
, = dt e e
T
t
t jn t jn
}
=

0
e e
= T dt dt e
T T
t n n j
} }
= =

0 0
) (
1
e

Chapter-2 14
) ( ), ( t u t u = dt
T

}
1
0
=T , the functions though orthogonal do not satisfy the requirement
of orthonormality. So we scale the functions to make them orthonormal among
themselves.
For both unit function and complex exponentials, scaling constant is T / 1 .
Therefore orthonormal bases are
{ } ,......... / 1 , / 1 , ,......... / 1 , / 1 , / 1
. . t jn t jn t j t j
e T e T e T e T T
e e e e

We denote the signal space generated by these bases as
2
L
Now any signal which is element of
2
L can be written as
...... / 1
/ 1 ...... / 1 / 1 ) ( / 1 ) (
.
. .
1
.
1 0
+ '
+ ' + + ' + ' + ' =

t jn
n
t jn
n
t j t j
e T C
e T C e T C e T C t u T C t f
e
e e e

(2.10)
As usual to get a particular coefficient, we project ) (t f on to the corresponding base.
}
= '
T
dt t u T t f C
0
0
) ( / 1 ) ( (2-11)
}

= '
T
t jn
n
dt e T t f C
0
.
/ 1 ) (
e
(2-12)
}

= '
T
t jn
n
dt e T t f C
0
.
/ 1 ) (
e
(2-13)
This result follows from the orthonormality of bases used in (2-10)
We now simplify the expression given in (2-10) by introducing new constants.
Let T C C / 1
0 0
' =
Since
}
= '
T
dt t u T t f C
0
0
) ( / 1 ) ( ,
} } }
= = =
T T T
dt t f
T
dt t u t f
T
dt t u T t f T C
0 0 0
0
) (
1
) ( ) (
1
) ( / 1 ) ( / 1 (2-14)
Similarly T C C
n n
/ 1 ' = =
}

T
t jn
dt e t f
T
0
) (
1
e
(2-15)
Chapter-2 15
T C C
n n
/ 1

' = =
}

T
t jn
dt e t f
T
0
) (
1
e
(2-16)
Thus we obtain complex fourier series in standard form
...... ...... ) (
. . .
1
.
1 0
+ + + + + + =

t jn
n
t jn
n
t j t j
e C e C e C e C C t f
e e e e
(2-17)
with the coefficients defined above in equations (2-14) to (2-16).

2.3 Summary
It is no understatement that the Fourier series especially Discrete Fourier Transform is
considered one of the crowning achievements of the 20th Century. The theory of Fourier
series has strong connection with geometry or at least we can understand the Fourier
Transform theory from a geometrical viewpoint. We had the honor and a privilege to
bring it to you, the reader, in this user friendly, low fat and no cholesterol form. In the
forth coming chapters on wavelets, this geometrical viewpoint will help you to digest the
theory in a better way. It will surely prevent intellectual diarrhea.

2.4 Exercises
1) Consider the interval [-1, 1] and the basis functions { } ,........ , , , 1
3 2
t t t for ] 1 , 1 [
2
L
The inner product on this vector space is defined as > < g f , =
}

1
1
. ) ( ) ( dt t g t f . You can
show that these functions do not form an orthonormal basis. Given a finite or countably
infinite set of linearly independent vectors
i
x , we can construct an orthonormal set
i
y
with the same span as
i
x as follows.

- Start with
1
y =
1
1
x
x

- Then, recursively set
2
y =
1 1 2 2
1 1 2 2
,
,
y y x x
y y x x
> <
> <

- Therefore,
k k
k k
k
v x
v x
y

= , where

=
> < =
1
1
,
k
i
i i k k
y y x v

This procedure is known as the Gram-Schmidt orthogonalization and it can be used to
obtain an orthonormal basis from any other basis. Follwing is a MATLAB function that
implement the Gram-Schmidt algorithm. Your input must be a set of linearly
Chapter-2 16
independent functions. The output will be a set of orthonormal functions that span the
same space.

function z=gramschmidt(x,N,M);
for i=1:N;
s(i,:)=x(i,:);
end;
e(1)=s(1,:)*conj(s(1,:).');
phi(1,:)=s(1,:)/sqrt(e(1));
for i=2:N;
th(i,:)=zeros(1,M);
for r=i-1:-1:1;
th(i,:)=th(i,:)+(s(i,:)*conj(phi(r,:).'))*phi(r,:);
end;
th(i,:)=s(i,:)-th(i,:);
e(i)=th(i,:)*conj(th(i,:).'); // note . means transpose without conjugation
phi(i,:)=th(i,:)/sqrt(e(i));
end;
z=phi(1:N,:);

Apply your algorithm on Legendre polynomials and plot the first six orhtonormal
basis functions you get.

2. Project the function f(x) = x onto the space spanned by ), 2 ( ), ( ), ( x x x |
] 1 , 0 [ ) 1 2 (
2
L x e where

s s
=
otherwise
x
x
, 0
2 / 1 0 , 1
) ( |

s s
s s
=
otherwise
x
x
x
0
1 2 / 1 1
2 / 1 0 1
) (
Sketch the projection of ) (x f to this space.

3. Show that the set of functions ), sin(
2
) ( mx x f
m
t
= m=1,2.. is an orthonormal
system in ] , 0 [
2
t L

4. Given that

(
(
(
(

(
(
(
(
(

(
(
(
(

(
(
(
(

=
2 / 1
2 / 1
0
0
,
0
0
2 / 1
2 / 1
,
2 / 1
2 / 1
2 / 1
2 / 1
,
2 / 1
2 / 1
2 / 1
2 / 1
4
H
Chapter-2 17
is an orthonormal basis for
4
9 and z=| |
E
z =
4
5
3
7
9
9 e
(
(
(
(

, where E is the standard basis for

4
9 . That is

(
(
(
(

(
(
(
(

(
(
(
(

(
(
(
(

=
1
0
0
0
,
0
1
0
0
,
0
0
1
0
,
0
0
0
1
E . Find z in terms of basis in
4
H

5. a)Find the fourier series of t t f = ) ( for t t < s t
b) Using Plancherels Formula and (a), evaluate the series

=1
2
1
n
n

[According to Plancherels Formula

e
=
Z n
n
C f
2 2
, where the
n
C s are complex fourier
series coefficients of the function ) (t f and
2
f in the given problem is dt t f
}

t
t
t
2
) (
2
1
]
6 Find the fourier series of

< s
< s +
=
t t
t t
x for
x for x
x f
0
0
) (


2.5 References

[1] Leon, Steven. Linear Algebra With Applications. Prentice Hall, New York, 1998.
[2] Strang, Gilbert. Introduction To Linear Algebra. Cambridge Press, Massachusetts,
1998.
[3] Burden, Richard L. Numerical Analysis. Brook/Cole, Albany, 1997.
[4] Ramirez, Robert W. The FFT: Fundamentals and Concepts. Prentice Hall, New York,
1985.
[5] Lyons, Richard. Understanding Digital Signal Processing. Addison Wesley, Menlo
Park, CA, 1998.
[6] Arnold, Dave. I Am The Resource For All That Is Math: A Contemporary Perspective
C.R., CA, 1998.