A frequency-domain implementation of the block P-vector adaptive algorithm

The P-vector approximate gradient algorithm is used in place of LMS when the desired signal is not available to the processor. This correspondence presents a frequency-domain implementation for use with single-input adaptive filters in which the coefficients are updated every L samples using the average gradient observed during the block. The development follows closely that presented previously for the LMS algorithm, and is slightly more general in that it is presented for complex time series.