Efficient implementation of error-feedback LSL algorithm

An error feedback LSL algorithm is presented which has a reduced computational complexity and makes use of both a priori and a posteriori prediction errors. A simple numerical convention is proposed to assure numerical stability for poor persistent excitation. Simulation results illustrate that divisions can be implemented with reduced wordlength without leading to a significant decrease in the overall accuracy even in floating-point implementations. This fact allows the use of lookup tables for implementing divisions. Moreover, the proposed algorithm has an inherent parallelism that can be advantageously exploited for fast implementations