Reduced-Memory Receiver Design for Adaptive Detection of Transient Signals

Abstract

Detection of one of M transient signals that recurs randomly in the present of noise is investigated. The signal is described as one of 2n fixed n-bit binary components with a time-recurrence structure that results in a signal ensemble of time sporadically recurrent binary components. The optimum likelihood-ratio detector can be realized as a sequential detector for this sporadic Poisson signal process. Even this sequential realization requires a rapidly expanding memory as the component ensemble size increases. This imposes a limitation on the practicality of implementation of the receiver design. Reduced-memory receiver designs are considered which provide near optimum detection performance. A memory scheme derived directly from the optimum detector design has been incorporated in a detector which exhibits very nearly optimum performance. This reduced-memory detector has been evaluated at various signal-to-noise ratios using the Receiver-Operating-Characteristic (ROC) curve. The reduction in memory afforded by the suboptimum detector appears to be quite significant, while variation from the optimum detector performance is relatively insignificant. [Work supported by the Office of Naval Research (Acoustic Programs).]