Improving the Speed of a Class of Single Frequency Estimation Algorithms

Recently, a class of recursive algorithms for single frequency estimation has been developed based on the rotate-add-decimate (RAD) method. The computational bottleneck of the RAD approach is the calculation of the arctangent function required to create a frequency estimate at the end of each recursion. This work demonstrates speed gains are possible with minimal loss in mean squared-error (MSE) performance by deferring accuracy in the calculation of the arctangent function until the final recursion.