A cumulant based TK method for the harmonic retrieval problem

The Tufts-Kumaresan (TK) method is well known as one of the best solutions to the harmonic retrieval problem. The fourth-order cumulant operator is used to modify the TK method. It is shown that, after applying the fourth-order cumulant operator to the observed signal, the form of the model of the harmonic retrieval problem still holds for the transformed data (cumulants of the observed signal). Therefore, the TK method is directly applicable to this new model for frequency estimation. The advantage of using the fourth-order cumulant operator is that the transformed data is insensitive to Gaussian noise (white or coloured), resulting in an effective increase of SNR. However, this is achieved at the expense of more computations. Computer simulation shows the effectiveness of this method