A block floating-point treatment to the LMS algorithm: efficient realization and a roundoff error analysis | IEEE Journals & Magazine | IEEE Xplore