Fast adaptation of frequency-domain volterra filters using inherent recursions of iterated coefficient updates

Adaptive Volterra filters are a popular model for compensating distortions caused by nonlinear structures with memory such as low-quality loudspeakers. This paper proposes a fast version of the recently investigated repeated coefficient updates for the partitioned block frequency-domain adaptive Volterra filter. Exploiting inherent recursions of the iteration procedure yields an efficient realization with a very low additional complexity compared to the usual LMS adaptation. Experimental results for both noise and speech demonstrate a significant acceleration of the filter convergence and overall echo cancellation for realistic nonlinear AEC scenarios.