It reduces to the standard Fourier uncertainty: if you have an energy packet (gaussian e.g.) of a certain width and wavelength, the shorter the width, the more uncertain your wavelength will be.
I think he means if you model things as waves the uncertainty comes about naturally. For example if you send a wave through a small hole so you know it's position accurately on two axes then that causes it to diffract out so you don't know it's direction/velocity along those axes.
