Speech Recognition for English Uttered by Japanese with Various Proficiency Levels
Hiroshi Kibishi, Seiichi Nakagawa · 2018
Recently, computer assisted language learning (CALL) systems have been subjected to extensive research. Particular focus has been placed on creating a voice-interactive CALL system that correctly recognizes spoken sentences to allow learners of a second language to get the most out of the system. In this paper, our aim is to create an English recognition system for Japanese speakers with various proficiency levels. We created various acoustic models using native and Japanese English data. Additionally, we proposed a deep neural network-hidden Markov (DNN-HMM) model and obtained improvement of absolutely 6% over using GMM-HMM.