Proof of the Kraft inequality for uniquely decodable codes: (a) (6 pts) Assume a uniquely decodable code has lengths
l_(1),dots,l_(M)
. In order to show that
\sum_j 2^(-l_(j))<=1
, demonstrate the following identity for each integer
n>=1
:
[\sum_(j=1)^M 2^(-l_(j))]^(n)=\sum_(j_(1)=1)^M \sum_(j_(2)=1)^M cdots\sum_(j_(n)=1)^M 2^(-(l_(j_(1))+l_(j_(2))+cdots+l_(j_(n))))
(b) (6 pts) Show that there is one term on the right for each concatenation of
n
codewords (i.e.. for the encoding of one
n
-tuple
x^(n)
) where
l_(j_(1))+l_(j_(2))+cdots+l_(j_(n))
is the aggregate length of that concatenation. (c) (6 pts) Let
A_(i)
be the number of concatenations which have overall length
i
and show that
[\sum_(j=1)^M 2^(-l_(j))]^(n)=\sum_(i=1)^(nl_(max)) A_(i)2^(-i)
(d) (6 pts) Using the unique decodability, upper bound each
A_(i)
and show that
[\sum_(j=1)^M 2^(-l_(j))]^(n)<=nl_(max)
(e) (6 pts) By taking the
n
th root and letting
n->\infty
, demonstrate the Kraft inequality.