Parameterization of the voice source by combining spectral decay and amplitude features of the glottal flow.