Masanori Kimoto

A stereophonic echo canceling algorithm for correct echo-path estimation

The authors had proposed a stereophonic echo canceller for correct echo path estimation. We pay attention to the convergence value of the filter coefficient vector of the multichannel least mean squares (MC-LMS) and that of the compact multichannel echo canceller (CEC) and use those differences. Concretely, the coefficient vector of the MC-LMS is converged to a correct echo-path by properly substituting a part of the coefficient vector of the CEC for that of the MC-LMS. We show by simulation for the white noise input that the coefficient vector of the proposed method converges to the correct echo-path.

A consideration on the modified block LMS-Newton algorithm and its performance

Adaptive algorithm especially plays the important role in adaptive signal processing, and various techniques has been proposed until up to now. In adaptive algorithms, both fast convergence characteristics and the reduction of the number of computational requirements per sample time are required. This paper presents a new block adaptive algorithm based on the projection onto one dimensional space as that is relatively satisfied with these requirements. Since the proposed algorithm makes its estimation quantity to distance of the coefficient vector in the unknown system and coefficient vector in the estimate system, step size of the proposed algorithm can be optimized to update coefficient vector of the estimate system.