Coding for Numerical Data Transmission

The usual approach to coding is to ignore the actual meaning of the transmitted symbols and to represent them in a purely statistical manner. As a result, all message errors are assumed to be equally costly and codes have been sought that simply reduce the probability that a message is received in error. While this may be appropriate for the transmission of some types of data, there are situations in which other criteria of goodness are of greater merit. If, for example, one is interested in the transmission of the temperature of a satellite, the probability that a particular observation is transmitted incorrectly may have little direct relation to system performance whereas a measure of the average magnitude by which the received data differ from the data actually transmitted could prove useful. * T h e material presented in this paper is based upon the dissertation, Coding for Numerical Data Transmission, submitted by the author to T h e Johns Hopkins University in conformity with the requirements for the degree Doctor of Philosophy. 1025