Signal detection using autoregressive parameters

The use of autoregressive parameter estimates is a well known technique for estimating the frequency of a sinusoid in white noise, whose phase and amplitude are perhaps also unknown. A more challenging problem is the problem associated with first detecting signal presence and then estimating one or more unknown parameters. This paper addresses the primary problem of detecting a signal, such as a sinusoid, with one or more unknown parameters, such as relative phase, frequency or bandwidth. The use of the autoregressive parameter estimation methods for detecting such signals is discussed, as well as the use of more conventional methods based on the Periodogram or other Fourier techniques. Performance comparisons are made based both on analysis and digital computer simulation.