摘要
文章在对应力影响下变异语音进行分频带分析的基础上,选用了可以提升语音信号中频段影响的修正Mel频率映射,并将其对应的MFCC系数作为新的语音识别特征。通过采用正常/变异语音分类器和新特征来进行变异语音识别。实验结果表明:采用修正Mel频率映射的MFCC特征改进了变异语音的识别性能。
On the basis of analyzing the recognition of speech under G-Force by using each Mel-frequency scale,two kinds of modified Mel-frequency scales which can emphasize the effect of the middle frequency ranges are explored in this paper,consequently,the relative MFCCs are selected as the features for recognition of speech under G-Force.Using a classifier of stressed speech and the new features, some experiments are conducted, and it shows that the modified Mel-frequency are effective for stressed speech recognition.
出处
《计算机工程与应用》
CSCD
北大核心
2002年第20期75-76,91,共3页
Computer Engineering and Applications
基金
国家自然科学基金(项目号:60085001)
教育部留学回国人员科研启动基金资助