Approximate processing for low-power digital filtering using variable canonic signed digit coefficients

An alternative approximate processing method which dynamically adjusts the precision of the canonic signed digit (CSD) coefficients is proposed. This proposed method employs a two-stage search for finding approximated values of CSD coefficients. Experimental results show that the number of additions is reduced to 29.4%, 11.8%, and 3.9% for each approximation level