No abstract

No abstract

Abstract. In this paper w e propose a batch learning algorithm for sequential blind extraction of arbitrary distributed but generally not i.i.d. (independent identically distributed) temporally correlated sources, possibly dependent speech signals from from linear mixture of them. The proposed algorithm is computationally very simple and efficient, it is based only on the second order statistics and in contrast to the most known algorithms developed for the sequential blind extraction and independent component analysis, do not assume statistical independence of source signals neither non-zero kurtosis for the sources, thus statistical dependent signals including sources with extremely low or even zero kurtosis (colored Gaussian with different spectra) can be also successfully extracted. Extensive computer simulation confirm the validity and high performance of the proposed algorithm.

We present a method to deal with adaptive noise cancelling based on independent component analysis (ICA). Although popular least-mean-squares (LMS) algorithm removes noise components based on second-order correlation, the proposed ICA-based algorithm can utilize higher-order statistics. Additionally, extending to transform-domain adaptive filtering (TDAF) methods, normalized ICA-based algorithm is derived to improve convergence rates. Experimental results show that the proposed ICA-based algorithm provides much better performances than conventional LMS approach in realworld problems.

The problem with the Hopfield associative-memory model caused by an imbalance between the number of ones and zeros in each stored vector is studied, and a modification of the Hopfield model that works well irrespective of the number of ones (or zeros) is proposed. This modified model can be implemented with no increase in memory.