﻿ Approximation of piecewise linear functions by discrete Fourier sums

Approximation of piecewise linear functions by discrete Fourier sums

DOI: 10.31029/demr.8.3

Let $N$ be a natural number greater than $1$. We select $N$ uniformly distributed points $t_k = 2\pi k / N$ $(0 \leq k \leq N - 1)$ on $[0,2\pi]$. Denote by $L_{n,N}(f)=L_{n,N}(f,x)$ $(1\leq n\leq N/2)$ the trigonometric polynomial of order $n$ possessing the least quadratic deviation from $f$ with respect to the system $\{t_k\}_{k=0}^{N-1}$. In the present article the problem of function approximation by the polynomials $L_{n,N}(f,x)$ is considered. Special attention is paid to approximation of $2\pi$-periodic functions $f_1$ and $f_2$ by the polynomials $L_{n,N}(f,x)$, where $f_1(x)=|x|$ and $f_2(x)=\mbox{sign\,} x$ for $x \in [-\pi,\pi]$. For the first function $f_1$ we show that instead of the estimate $\left|f_{1}(x)-L_{n,N}(f_{1},x)\right| \leq c\ln n/n$ which follows from well-known Lebesgue inequality for the polynomials $L_{n,N}(f,x)$ we found an exact order estimate $\left|f_{1}(x)-L_{n,N}(f_{1},x)\right| \leq c/n$ ($x \in \mathbb{R}$) which is uniform with respect to $1 \leq n \leq N/2$. Moreover, we found a local estimate $\left|f_{1}(x)-L_{n,N}(f_{1},x)\right| \leq c(\varepsilon)/n^2$ ($\left|x - \pi k\right| \geq \varepsilon$) which is also uniform with respect to $1 \leq n \leq N/2$. For the second function $f_2$ we found only a local estimate $\left|f_{2}(x)-L_{n,N}(f_{2},x)\right| \leq c(\varepsilon)/n$ ($\left|x - \pi k\right| \geq \varepsilon$) which is uniform with respect to $1 \leq n \leq N/2$. The proofs of these estimations based on comparing of approximating properties of discrete and continuous finite Fourier series.

Keywords: function approximation, trigonometric polynomials, Fourier series.

﻿