A method for the practical evaluation of the Hilbert transform on the real line (Q1964075)

The author presents a method for the numerical evaluation of the Hilbert transform over the entire real line \[ H[f](t)= \int^{+\infty}_{-\infty} \exp(-x^2) f(x)(x- t)^{-1} dx,\quad t\in\mathbb{R}, \] with the Hermite weight function and where the integral is taken in the Cauchy principal value sense. The method is similar to the one described for the finite interval case by \textit{G. Mastroianni} and \textit{D. Occorsio} [Comput. Math. Appl. 30, No. 3-6, 155-168 (1995; Zbl 0837.41002)] and can be also applied to other weight functions. It consists of a combination of two polynomial approximation processes. The main parts of the calculation procedure are concerned with the underlying quadrature formula and polynomial interpolation. The final approximation for \(H[f](t)\) is given by \[ H_n[f](t)= \sum^{n+ 1}_{k= 1}\widehat H_n[f](x_{j, n+1}) l_{k,n+1}(t), \] where \(\widehat H_n[f](t)\) involves the \(n\)-point Gauss-Hermite quadratic formula and \(l_{k,n+1}\) is the \(k\)th fundamental polynomial of the Lagrange interpolation process associated with the zeros \(x_{j,n+1}\), \(j= 1,2,\dots, n+1\) of the \((n+1)\)th Hermite polynomial. The above procedure is numerically stable for all \(t\), and computionally efficient. A short error analysis for the method is made and some numerical examples are finally given.