Noise cancellation in sub-bands using the conjugate gradient method

An adaptive filter bank is considered for noise cancellation with the conjugate gradient (CG) adaptive algorithm being used as the optimization tool. It uses the exponentially windowed pass instantaneous gradients and it has convergence properties superior to the LMS and a lower computational complexity than the RLS. With the appropriately designed analysis and synthesis filter banks and the decimation factor such that aliasing can be ignored, the proposed technique provides satisfactory performance with faster convergence and high computational efficiency. Therefore, it is an attractive alternative for real time noise cancellation