This research seeks to develop the theory and practice of an integrated signal detection system for detection in a narrowband burst environment such as is encountered in sonar, radar, data communications, and other similar environments. Optimal solutions to signal detection problems involve complete statistical descriptions of the ambient noise. However, many important environments are non- stationary, the noise consisting of a low-power quiescent-mode random process punctuated occasionally by higher-power bursts. In some environments, these bursts are strongly narrowband, a property that must be considered. This research will be in three parts: characterization and modelling of bursts, the development of processing algorithms, and investigation of the properties of the algorithms using both idealized simulation and actual data.