Predicting human perception and ASR classification of word-final [t] by its acoustic sub-segmental properties