Text this: Vowel recognition based on acoustic and visual features