Optimizing computations in a transposed direct form realization of Floating-Point LTI-FIR systems

Abstract

The inherent computational redundancy in discrete-time LTI-FIR system response computations in Digital Signal Processing have been exploited in a variety of ways to minimize the computational complexity. We present an improved algorithm-level computational optimization that uses sorted recursive differences between coefficients representing the system transfer function with a Floating-Point number representation to extract maximum benefits from this redundancy. A can be applied to any LTI-FIR system and there is no deterioration in accuracy compared 20 directly using the coefficients. The results for several practical FIR systems show that there is a significant reduction in the computational complexity, hence power consumed, using this technique.