论文信息 - Design and development of an objective English accent recognition system based on RankNet's word stress recognition method

Design and development of an objective English accent recognition system based on RankNet's word stress recognition method

In recent years, speech processing technology has been widely used in the field of auxiliary language learning. It provides an auxiliary learning platform for non-native language learners. This paper develops an objective English accent recognition system to help English learners grasp the rhythm of spoken English. On the basis of RankNet's word stress recognition method, local normalization is added to improve the recognition accuracy of RankNet. We further apply RankNet to sentence recognition. According to the RankNet method, the rhythm characteristics of non-specific people and the quality of vowel phonemes are selected to classify them into the stressed and unstressed, and the accuracy is improved according to the result of prosodic recognition of statements.

Qi Wang | Yonghuai Zhang

[1] Su Fei,et al. Progressive framework for deep neural networks: from linear to non-linear , 2016 .

[2] Zhu Qi,et al. Multi-radio access based bandwidth allocation strategy for video communication in heterogeneous wireless networks , 2016 .

[3] Chong-Jia Ni,et al. Mandarin Stress Detection Using Acoustic, Lexical and Syntactic Features: Mandarin Stress Detection Using Acoustic, Lexical and Syntactic Features , 2011 .