Digital Signal Processing Without Arithmetic Using Regression Trees
Jake Gunther, Todd K. Moon · 2009
This paper discusses the possibility of performing digital signal processing tasks such as filtering without doing arithmetic, i. e. addition and multiplication, by relying on the approximating capabilities of regression trees. Regression trees map a vector of inputs to a scalar output using only comparison (greater than, less than) operations. Therefore, regression trees can approximate the value of a given function without doing arithmetic offering a computational savings. This paper demonstrates the performance of regression trees on simple FIR filtering problems.