The Bedrosian identity and homogeneous semi-convolution equations (Q1001612)

If for two functions \(f,g\in L^2(\mathbb R)\), \(\text{supp}\hat{f}\subset[0,\infty)\) \(\text{supp}\hat{g}\subset[0,\infty)\) or \(\text{supp}\hat{f}\subset[-a,a]\) and \(\text{supp}\hat{g}\subset(-\infty,-a]\cup[a,\infty)\) holds, then there holds the formula \[ [H(fg)](x) = f(x) (Hg)(x),\quad\text{almost everwhere} \quad x\in\mathbb R\tag{1} \] which is called the Bedrosian identity, where \(\text{supp}\hat{f}\) denotes the support of the Fourier transform \(\hat{f}\) of \(f\). The circular Bedrosian identity is defined by \[ [\tilde{H}(fg)](t)=f(t)(\tilde{H}g)(t),\quad\text{almost everwhere}\quad t\in[0,2\pi],\tag{2} \] where \(\tilde{H}\) is the circular Hilbert transform defined for each \(f\in L_{2\pi}^1\) at \(t\in[0,2\pi]\). It is shown by the authors that the Bedrosian identity defined by (1) is equivalent to two homogeneous semi-convolution equations. In order to investigate these equations, the characterizations of right translation invariant subspaces of \(L^2(\mathbb R_+)\) and additive positive definite kernels on \(\mathbb R_+\) are utilized. The results obtained by the authors are slight improvement over the results given earlier by \textit{Y. Xu} and \textit{D. Yan} [Proc. Am. Math. Soc. 134, No. 9, 2719--2728 (2006; Zbl 1101.44002)] and by \textit{T. Qian, Q. Chen} and \textit{L. Li} [Physica D 203, No. 1--2, 80--87 (2005; Zbl 1070.94504)]. Finally the authors construct intrinsic mode functions by making use of the results established in the paper.