The classification of Mandarin finals using distinctive features

Distinctive features are the elemental units of phonology which govern the behavior of segments within language. It is of great significance to speech recognition, synthesis and pronunciation scoring. This paper study the acoustic cues of distinctive features based on the distinctive features framework of mandarin finals proposed by Zhang jialu. Then Classifiers of Mandarin finals were constructed based on distinctive features. At first, acoustic cues of distinctive features were extracted and two-class classifiers were constructed by support vector machine (SVM) to check the binary values of distinctive features. After that, each final was identified according to a specific binary search process in distinctive feature trees and multi-level classification models of Mandarin finals were constructed.