Improving the threshold of the minimum-norm frequency detection method

This work is concerned with the frequency estimation of sinusoidal signals at low signal-to-noise ratios (SNR) and high resolution, through the linear prediction (LP) minimum-norm-based method. The optimized LP coefficients lead to a polynomial whose zeros contain information about the frequencies. The study of the behavior of these zeros leads to a new strategy capable of lowering the threshold effect at low SNR. Simulations show that is possible to achieve better results than the maximum likelihood (ML) method without any significant additional effort.