Automatic pronunciation assessment for language learners with acoustic-phonetic features

摘要

Computer-aided spoken language learning has been an important area of research. The assessment of a learner's pronunciation with respect to native pronunciation lends itself to automation using speech recognition technology. However phone recognition accuracies achievable in state-of-the-art automatic speech recognition systems make their direct application challenging. In this work, linguistic knowledge and the knowledge of speech production are incorporated to obtain a system that discriminates clearly between native and non-native speech. Experimental results on aspirated consonants of Hindi by 10 speakers shows that acoustic-phonetic features outperform traditional cepstral features in a statistical likelihood based assessment of pronunciation.