IIR notch filtering-comparisons of four adaptive algorithms for frequency estimation

This paper compares the parameter estimation accuracy of four adaptive algorithms for frequency estimation when applied to an IIR digital notch filter. All four algorithms were subjected to the same experimental conditions and the variance of parameter estimates are compared to the Cramer Rao Lower Bound. Results show that the RML yielded the most accurate parameter estimates although its computational burden is quite high. The AML produced good parameter estimates and it has the advantages of proven convergence properties as well as lower computational burden over the RML. For applications where the signal to noise ratio is moderate it is shown that the AGB algorithm may be suitable, particularly where minimal computational burden is desired.