-
all-pole model of linear predictor leads to a limitation in the analysis of nasals such as /n/ in linear, where transfer function zeros occur in the model with a nasal branch, the utterance of such words also cause adjacent vowels to be nasalized
-
the major disadvantage of the LP model in speech analysis is that all-pole model A(w) approximates the short-term power spectrum P(w) equally well at all frequencies of analysis band, which is inconsistent with human hearing because of it's non-linear nature above 800 Hz