06.06.2022 Views

B. P. Lathi, Zhi Ding - Modern Digital and Analog Communication Systems-Oxford University Press (2009)

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

l 0. 10 Other Useful Performance Criteria 579

and the decision rule in Eq. (10.134) reduces to the rule in Eq. (10.83), as expected. The

generalized Bayes receiver for M = 2, assuming C1 1 = C22 = 0, sets m = mi if

Otherwise, the receiver decides that m = mz.

Maximum Likelihood Receiver

The strategy used in the Bayes receiver discussed in the preceding subsection is general, except

that it can be implemented only when the a priori probabilities P(m1 ), P(m2), ... , P(mM) are

known. Frequently this information is not available. Under these conditions various possibilities

exist, depending on the assumptions made. When, for example, there is no reason to expect

any one signal to be more likely than any other, we may assign equal probabilities to all the

messages:

1

P(m1 ) = P(m2) = · · · = P(mM ) = - M

Bayes' rule [Eq. (10.83)] in this case becomes: set m = mk if

for all i i= k (10.136)

Observe that pq (qlmk) represents the probability of observing q when mk is transmitted. Thus,

the receiver chooses that signal which, when transmitted, will maximize the likelihood (probability)

of observing the received q. Hence, this receiver is called the maximum likelihood

receiver. Note that the maximum likelihood receiver is a Bayes receiver for the cost of

Eq. (10.135) under the condition that the a priori message probabilities are equal. In terms

of geometrical concepts, the maximum likelihood receiver decides in favor of that signal

which is closest to the received data q. The practical implementation of the maximum likelihood

receiver is the same as that of the Bayes receiver (Figs. 10.18 and 10.19) under the

condition that all a priori probabilities are equal to 1 / M .

If the signal set is geometrically symmetrical, and if all a priori probabilities are equal

(maximum likelihood receiver), then the decision regions for various signals are congruent.

In this case, because of symmetry, the conditional probability of a correct decision is the same

no matter which signal is transmitted, that is,

P(Clmi) = constant

for all i

Because

M

P(C) = L P(mi)P(Clmi)

i=l

in this case

P(C) = P(Clmi) (10.137)

Thus, the error probability of the maximum likelihood receiver is independent of the actual

source statistics P(mi) for the case of symmetrical signal sets. It should, however, be realized

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!