Speaker-independent phonetic classification using hidden Markov models with mixtures of trend functions | IEEE Journals & Magazine | IEEE Xplore