Accounting for measurement noise color in frequency domain adaptive algorithms

A method of using a partial knowledge of measurement noise in frequency domain adaptive algorithms is considered. Overlap-save and overlap-add type adaptive schemes are both investigated. The algorithms are derived by minimizing a best linear unbiased estimators (BLUE) criterion function using the stochastic gradient method and then switching over to the frequency domain to reduce the computational complexity. The performance of the algorithms is investigated via simulations using stationary as well as speech signals.