Frequency and Magnitude Response Design Approximation Problems for Nonlinear-phase Nonrecursive Digital Filters (Q4782602)

The paper deals with an important problem arising in Digital Signal Processing (DSP), namely the design of Finite Impulse Response (FIR) filters whose frequency response reads as follows \[ H_N(x, \omega) =\sum^{N-1}_{k=0} x_ke^{-jk\omega},\;\omega\in\mathbb{R}. \] The set \([x_0,x_1, \dots,x_{N-1}]\) constitutes the filter coefficient vector \(x^T\) whereas \(N\) is called the filter length. Usually, \(x\) is a real-valued vector, but complex-valued is also allowed. In applications, \(N\) is usually 200--300. The problem is to find \(x\) so that \(H_N\) be as near as possible to a given (complex-) valued continuous function \(D\), called the desired frequency response and defined on \([0,\pi]\) or \([-\pi,\pi]\). The nearness is measured by some norm of the difference between \(D\) and \(H_N\). So, it is of interest to know the conditions under which the error tends to zero when \(N\) tends to infinity. Four basic problems arising in FIR filter design are correctly stated in the frame of mathematics. Some basic results are given which insure the approximation convergence. At the end of the paper some considerations related to numerical procedures used in FIR filter design are given. Although the paper gives no numerical examples, it provides the mathematical foundations of some procedures heuristically proven in the classical literature in DSP. So the paper is of high interest for mathematicians working in approximation theory as well as for practitiones in FIR filter design theory.