Assessment of cumulant-based approaches to harmonic retrieval

The authors answer the following questions regarding solving the harmonic retrieval problem using second- or fourth-order statistics: (1) How do cumulant-based results compare against correlation-based results?; (2) How robust are the cumulant-based methods to tones of different local SNRs?; (3) How close can the two tones be brought together before cumulant-based methods break down?; (4) Does one have to use a high-resolution method, such as MUSIC or Minimum Norm, or can the Pisarenko method be used?; and (5) What is the 'footprint' of success for the cumulant-based methods? Extensive simulations are used to provide the answers.<
>