摘要
针对目前语言学习中普遍存在的发音问题,提出了一种基于MFCC的语音评分方法,通过对测试语言和标准语音进行预处理、特征提取及模式比对等过程,找出它们的相似度并由评分机制给出得分。实验证明该方法能够确保评测的准确性,具有较强的实用性。
In order to solve the common inaccurate pronunciation problems,the paper proposes an algorithm for accomplishing speech evaluation based on MFCC.It takes pre-processing,feature extraction and pattern matching to the test speech and the standard one and tries to find the similarity between them.The result is given by the mechanism.Experimental results show that the algorithm which retains the comparable performance has a strong practicality.
出处
《中国现代教育装备》
2010年第15期41-44,共4页
China Modern Educational Equipment
关键词
语音评分
特征提取
动态时间规整法
梅尔频率倒谱系数
speech evaluation
feature extraction
dynamic time warping
mel-frequency cepstral coefficients（MFCC）
