An Adaptive Blind Separation Algorithm for Co-frequency Overlapping Signals: RLS-based Approach

This paper presents a novel adaptive blind separation algorithm for co-frequency overlapping signals, identified as the RLS-based adaptive blind separation algorithm. The primary objective of this algorithm is to separate co-frequency overlapping signal components from received signals without requiring any prior information about these components. The distinctive feature of this algorithm lies in its implementation of the Recursive Least Squares method to estimate the mixing matrix of signal components, enabling blind signal separation through matrix inversion techniques. The algorithm employs real-time weight updates using the RLS filter, which minimizes the mean square error through recursive computation of the inverse correlation matrix. Implementation typically involves initializing filter weights, calculating the gain vector, and updating weights iteratively for each new sample. Furthermore, the algorithm demonstrates adaptive capabilities that automatically adjust to signal variations, thereby achieving superior separation performance under dynamic conditions. The RLS implementation ensures faster convergence compared to LMS-based approaches while maintaining computational efficiency through matrix inversion lemma applications. Overall, this RLS-based adaptive blind separation algorithm provides an effective solution for separating co-frequency overlapping signals in practical communication systems.