Efficient allocation of power of two terms in FIR digital filter design using tabu search

This paper concerns the design of FIR filters using quantized coefficients belonging to a sum of signed power-of-two terms (SPT) domain. Unlike other SPT filter design approaches, we added a global constraint that a priori fixes the total number of shift registers; thus each coefficient can be represented using different precision. The difficult FIR filter design optimization task, is solved by a specific Tabu Search (TS) method. Several experimental results and comparisons with other well known approaches, demonstrate the effectiveness of this filter architecture and the proposed optimization algorithm