Blind Separation Algorithm with Improved Learning Rule for Audio Signal

In this paper, the separation model for audio signal is built, at the same time, separable assumption conditions and separation criteria are assumed. Improved learning rule is proposed, on the basis of the separation effect with two learning rules, and it is applied in the adaptive algorithm to realize blind separation for audio signals. Experiment shows that the performance improved algorithm is better in the speed of convergence is fast, stability of separation signals and separation effect than the two original learning rules. Moreover it eliminates the phenomenon that the step decreases too late or too early.