Low-complexity detection of small frequency deviation by the generalized LMPU test

In this paper, we consider the detection of a small deviation of the frequency of sinusoidal signals, which arises in various signal processing applications. The generalized likelihood ratio test (GLRT) for this problem uses the maximum likelihood (ML) estimator of the frequency, and therefore suffers from high computational complexity. In addition, the GLRT is not necessarily optimal and its performance may degrade for non-asymptotic scenarios that are characterized by close hypotheses and small sample sizes. In this paper we propose a new detection method, named the generalized locally most powerful unbiased (GLMPU) test, which is a general method for local detection in the presence of nuisance parameters. A closed-form expression of the GLMPU test is developed for the detection of frequency deviation in the case where the complex amplitudes of the measured signals and the noise variance are unknown. Numerical simulations show improved performance over the GLRT in terms of probability of detection performance and computational complexity.