基于特征的藏文音节识别算法  

Tibetan syllable recognition algorithm based on the features

作　　者：张日培 姜占才[1] ZHANG Ri-pei;JIANG Zhan-cai(School of Computer Science,Qinghai Normal University,Xining 810008,China)

机构地区：[1]青海师范大学计算机学院,青海西宁810008

出　　处：《电子设计工程》2018年第20期137-142,147,共7页Electronic Design Engineering

基　　金：国家社会科学基金(15XYY026)

摘　　要：为了实现藏文的文语转换(TTS),提出基于字符投影变换特征的藏文音节识别算法。该算法以音节为基元,选择并提取音节中由字符列投影变换组成的特征向量,以此建立音节特征库;通过查表算法对藏文音节进行识别。算法还包括藏文文本的规范化和音节切分两部分内容。通过理论分析和算法测试实验证明:提取的特征向量与藏文音节一一对应,藏文音节识别率达到100%,且特征的提取过程简便易行。该算法已经成功应用于藏文的文语转换系统。For the purpose of realizing the translation of Tibetan language （TTS） , a Tibetan syllable recognition algorithm based on character projection transformation is proposed. The algorithm is on the basis of syllable, selects and extracts feature vectors formed by projection transformation in the character syllables in the hope of setting up syllable feature bank, and it also identifies Tibetan syllable by the method of look-up table algorithm. The algorithm includes two parts： the normalization of Tibetan text and the segmentation of syllables. Theoretical analysis and algorithm test has proved that the extracted feature vectors are each corresponded to Tibetan syllables, and the rate of Tibetan syllable recognition reaches 100%. The process of feature extraction is simple and feasible. The algorithm has been successfully applied to the Tibetan language translation system.

关 键 词：计算机应用技术 藏文文语转换 模式识别 音节识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]