A unified approach to split structure adaptive filtering

In this paper, a pragmatic approach to continuously split up an adaptive transversal filter is developed. It is shown that any M=2 ^L length adaptive filter can be decomposed into M parallel single-parameter sub-sections by an L-step splitting operation. The sub-sections can then be adapted separately and the convergence behaviour of the overall system is greatly improved. The proposed algorithm has a simple modular structure and its speed of convergence is much faster than the conventional method. A unified perspective between Walsh transform adaptation and split structure adaptive filtering is also illustrated.