Sequential Detection of a Narrowband Signal Emerging in Noise (SekventiellDetektering av en Smalbandig Signal Uppdykande i Brus)

摘要

Sequential hypotheses-testing techniques are useful for derivation and analysisof algorithms aimed at detecting sudden changes in the state of various signal processing systems. We, in this report propose a change detector that is applicable under general circumstances. Specifically, we allow for time-varying distributions with unknown parameters and dependent samples. The derivation of the detector is motivated from the structure of the CUSUM algorithm, and consequently it belongs to the same family of procedures. We also point out that this detector specializes to a couple of other ones proposed in the literature under less general assumptions than the ones we have made. A computer simulation study is conducted to illustrate the performance of the detector in the case of a sinusoidal signal with known frequency and unknown amplitude and phase emerging in white Gaussian noise.