Efficient Design of Sparse Multiplierless FIR Filters with Low Complexity

Abstract

FIR filters have many advantages such as linear phase, high precision and good stability. However, when the performance is required to be high, usually a higher order is required, resulting in greatly increased hardware complexity of the FIR filter. Based on sparse FIR filter design algorithm and common subexpression elimination method, a novel algorithm to design the FIR filters with low complexity. First, a sparse FIR benchmark filter that fulfills frequency response specifications is yielded from the sparse filters design algorithm. Then, each quantized filter coefficient is represented in CSD. And the weight of all subexpressions and isolated nonzero digits of the quantized coefficient set are computed. Finally, the filter coefficient set with lower implementation cost is constructed by iteratively admitting subexpressions and isolated nonzero digits according to their weight. The simulation results show that the proposed algorithm can saves about 29% of adder compared with other low complexity filter design algorithms, which effectively reduces the implementation complexity and greatly saves the hardware cost.