A high-throughput DLMS adaptive algorithm

The high-throughput delayed LMS (DLMS) adaptive algorithm suffers from a slower convergence rate compared to the LMS algorithm. Different versions of the DLMS adaptive algorithm using a conversion scheme have been proposed to improve the convergence rate. This improved convergence was achieved at the expense of an increased computational complexity and a lower throughput rate than the original DLMS algorithm. We propose a new modified DLMS adaptive algorithm that, compared to the existing conversion-based DLMS algorithm, provides a higher throughput rate for a similar convergence rate. Alternatively, the proposed algorithm provides a faster convergence for the same throughput rate compared to the conversion-based DLMS algorithm. In both the cases, the computational complexity of the proposed algorithm is smaller than that of the conversion-based DLMS algorithm. The proposed algorithm uses the error signal from each stage of the adaptive FIR filter independently to update the value of the corresponding coefficient. Simulations illustrate the convergence performance of the new algorithm. The performance of its architecture is evaluated in terms of computational complexity, throughput, and latency. The proposed algorithm provides a better throughput rate and a computational complexity lower than that of the conversion-based DLMS algorithm.