论文部分内容阅读
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。
This paper presents a method for automatic detection and evaluation of children’s melody in Chinese Mandarin test. Under the existing computer aided speech evaluation system, the pronunciation and acoustic characteristics of melodious phonology are deeply analyzed, and the detection and evaluation of phoneticized phonetics are transformed into typical classification problems. After selecting a number of representative acoustic features and attempting a variety of different classification algorithms, the results show that the Boosting CART enhanced classification model can fully utilize the various acoustic features, Reached 92.41%. Through the further analysis of the acoustic feature set, it is found that the resonance peaks, the confidence of the pronunciation and the duration are the most important clues for expressing the murmur. By using these clues, the automatic detection and evaluation of the murmur can be realized effectively.