汉语普通话水平测试中儿化音的自动检测与评价

发布时间：2018-05-02 14:09

本文选题：普通话水平测试 + 声学特征　；参考：《声学学报》2014年05期

【摘要】：提出一种汉语普通话水平测试中儿化音的自动检测与评价方法。在现有计算机辅助发音评测系统的框架下,深入分析儿化音的发音规律和声学特性,将儿化音的检测与评价转化成典型的分类问题进行处理。经过挑选多个有代表性的声学特征,并尝试多种不同的分类算法,结果表明,集成分类回归树(Boosting CART)强化分类模型,能充分利用儿化音的各种声学特征,分类正确率达到92.41%。通过对声学特征组的进一步分析,发现共振峰、发音置信度、时长是表达儿化音的最重要线索,利用这些线索能有效地实现对儿化音的自动检测与评价。
[Abstract]:This paper presents a method for automatic detection and evaluation of Chinese Putonghua pronunciation in Chinese Putonghua proficiency test. In the framework of the existing computer-aided pronunciation evaluation system, the phonetic rules and acoustic characteristics of the phonetics are deeply analyzed, and the detection and evaluation of the phonetics are transformed into typical classification problems to be dealt with. By selecting several representative acoustic features and trying different classification algorithms, the results show that the improved classification model can make full use of the acoustical features of the catechu sound, and the classification accuracy is 92.41%. Through further analysis of acoustic feature group, it is found that resonance peak, confidence degree of pronunciation and duration are the most important cues to express the catechu sound. The use of these cues can effectively realize the automatic detection and evaluation of the phonetics.
【作者单位】：哈尔滨工业大学计算机科学与技术学院;哈尔滨师范大学计算机科学与信息工程学院;
【基金】：国家自然科学基金(61171186) 黑龙江省自然科学基金(F201321) 黑龙江省应用技术研究与开发计划项目(GZ13A003)资助
【分类号】：TN912.3

【相似文献】