Академический Документы
Профессиональный Документы
Культура Документы
Example:
To find gcd(x5 + x4 + x + 1, x4 + x2 + 1), do
Euclid with these two polynomials as inputs
(x5 + x4 + x + 1) (x + 1) (x4 + x2 + 1) = x3 + x2
(x4 + x2 + 1) (x + 1) (x3 + x2 ) = 1
(x3 + x2 ) (x3 + x2 ) (1) = 0
So the gcd of x5 + x4 + x + 1 and x4 + x2 + 1 is
1.
4
Example:
To find gcd(x6 + x5 + x3 + x + 1, x4 + x2 + 1),
do Euclid with these two polynomials as inputs
(x6 +x5 +x3 +x+1)(x2 +x+1)(x4 +x2 +1) = 0
Thus, the gcd of x6 + x5 + x3 + x + 1 and
x4 + x2 + 1 is x4 + x2 + 1. (Not at all obvious
that the second poly divides the first!)
Example: To find
gcd(x6 + x5 + x3 + x + 1, x4 + x3 + x + 1), do
Euclid with these two polynomials as inputs
(x6 +x5 +x3 +x+1)x2 (x4 +x3 +x+1) = x2 +x+1
(x4 + x3 + x + 1) (x2 + 1) (x2 + x + 1) = 0
Thus, the gcd of x6 + x5 + x3 + x + 1 and
x4 + x3 + x + 1 is x2 + x + 1.
= 1
= 0
(1)(x2 + x + 1) (x + 1)(x)
Since in general
1=rf +sg
implies that r is a multiplicative inverse of f
mod g we see that x + 1 is a multiplicative
inverse of x mod x2 + x + 1.
=
(1)(x2 + 1) + (x)(x)
= (x2 +1)+(x)((x3 +x2 +1)+(x+1)(x2 +1))
= (x)(x3 + x2 + 1) + (x2 + x + 1)(x2 + 1)
Since in general
1=rf +sg
implies that r is a multiplicative inverse of f
mod g, x2 + x + 1 is a multiplicative inverse of
x2 + 1 mod x3 + x2 + 1.
Back to codes
Review:
Memoryless, binary, discrete channels
We always do minimum-distance decoding
This is the same as max likelihood decoding
This includes error correction: If a received
word is closer to the sent word than to any
other codeword, the correction is correct.
If by mischance there are so many bit
errors that the received word is closer to a
different codeword than the one sent, then the
correction is wrong.
... but we have no way of knowing this.
As a default, we imagine that any pattern
of errors with more bit errors than half
the minimum distance between codewords
will not be correctly corrected.
If the minimum distance is d = 2e + 1, then any
e bit errors can be (correctly) corrected.
10
Hamming bound
Using the physical analogy that Hamming
distance is really like distance:
the set of all length n codewords with an
alphabet with q letters (maybe q = 2) is like
a container
codewords with specified minimum distance
d = 2e + 1 between them are like balls of radius
e
and the question of how many codewords
of length n (alphabet size q) with minimum
distance d = 2e + 1 can be chosen is analogous
to asking
How many balls of a fixed radius can be packed
into a box with a specified volume?
This is a hard question, but an easier version is
definitive:
the total volume of the balls packed cannot be
greater than the volume of the container.
(Duh!)
12
=
=
1
n
1
n
2
=
...
=
n
3
(q 1)
(q 1)2
(q 1)3
(q 1)e
n
e
13
n
n
n
q ` 1+
(q 1) + . . . +
(q 1)e
1
e
This is the Hamming bound.
If a code exists with ` codewords, of length n,
and minimum distance d = 2e+1, this inequality
must hold.
The contrapositive assertion is that, given `, n,
q, and d = 2e + 1 if the inequality fails then
there cannot exist any such code.
14