Incorporating FIR Filter using MIP Algorithm in Denoising ECG Signal

Transposed direct-form FIR filters which are more effective and realizable structure than direct form are implemented by multiple constant multiplications (MCM) architecture. Early works have focused on replacing MCM blocks by simple operators such as addition, subtraction and shift to form adder-tree. As the complexity of such filters is dominated by the number of additions/subtractions in the adder-tree, common sub expression elimination (CSE) is used by sharing common terms across all multiplications. A novel algorithm in the scheduling of adder-tree operations called mixed integer programming (MIP) algorithm is investigated to enable exact bit-level optimization of adder-trees for efficient FIR filter implementation.

Experimental results shows that upto 4.7% of average area reduction and 3.6% of average power reduction can be achieved on the optimized adder/subtractor tree of the MCM block.