On detection and maximum likelihood estimation of multiple sinusoids in white noise

摘要

The paper presents a new computationally efficient algorithm for estimating the number and the parameters of multiple closely spaced real tones (sinusoids) in noise from a block of T uniformly spaced samples. It consists of a combination of a robust statistics to detect both the number of sinusoids as well as to deliver initial estimates of their frequencies, and of a direct maximum likelihood parameters estimator.