A stable algorithm for block adaptive signal processing

The authors present a block time-domain adaptive algorithm which is equivalent to the conjugate-gradient algorithm for solving linear equations, where minimizing block-by-block squared-error sum leads to the linear equations. The proposed algorithm can provide an efficient solution for adaptive signal-processing applications, in which the number of available data is less than that of the filter coefficients or data are discontinuously available, and for those in a noisy environment. Experimental results indicate that the algorithm converges in a single block without creating an unstable performance caused by the presence of quantization error, finite-precision error, and additive noise.<<ETX>>