Adaptive noise cancellation using fast optimum block algorithms

The application of block frequency domain (FFT-based) adaptive algorithms to noise cancellation is considered. Two classes of FFT-based algorithms are examined, namely those associated with periodic convolution and those associated with linear convolution. In both cases normalized optimum convergence factors are considered. Experimental noise cancellation results are given using speech corrupted by white, colored, and harmonically structured noise (simulated engine noise). Quantitative and subjective evaluations are given for all the results, and trade-offs of performance versus computational complexity are established