All-Pole Spectral Modelling Of Voiced Speech With A Highly Compressed Set Of Parameters