Tracking Formants, Extra-Formants and Anti-Formants in Non-Modal Speech by Means of a Spectral Pole-Zero Model