Comparative Evaluation of Speech Parameterizations for Speech Recognition