Characterizing the Hilbert transform by the Bedrosian theorem