THE FORMULATION AND ANALYSIS OF THE THEORY FOR DETERMINING REQUIRED RECEIVER MEMORY FOR SIGNAL DETECTION.
DUKE UNIV DURHAM N C ADAPTIVE SIGNAL DETECTION LAB
Pagination or Media Count:
The purpose of the report is to develop and evaluate a theory of memory applicable in statistical detection theory. An optimum finite memory approach is postulated and the resulting signal detector designs are evaluated for a class of problems including Signal-Known-Exactly SKE, Signal-Known-Except Amplitude SKEA, and M-ary Signaling. The finite memory detector design for a class of problems involving a transient signal is considered also. The transient signal is described in terms of one of a finite number of possible time varying waveforms which recur synchronously. The fixed-ended finite memory design for the non-sequential SKE problem is presented. This design involves problems where the observation length is specified a priori and a decision output is rendered only at the end of the observation. Also an open-ended finite memory design is presented which results from assuming a priori neither the eventual length of obervation nor the maximum length of observation. Using the Receiver Operating Characteristic ROC the open-ended finite memory design is evaluated and compared with the optimum infinite memory design. Results show that the infinite memory detector performance can be obtained by a very small finite memory detector involving seven states. The trade-off between detector memory and time is also apparent. Author