摘要
为实现针对课堂教学语音数据内容的准确分类,提出一套基于GMM和UBM的双重聚类发言者分类识别算法。根据课堂教学活动师生发言不均衡的特点,制定专门的双重聚类处理方案,并通过课堂发言实录语音数据对该算法进行应用实验。经实验研究发现,所提出的GMM-UBM的双重聚类算法相比于单纯GMM来说能够更加准确地对发言者进行识别与分类,具有一定的应用价值。
In order to realize the accurate classification of the content of classroom teaching speech data,this study proposes a set of dual clustering recognition algorithm based on GMM and UBM.According to the characteristics of unbalanced speech between teachers and students in classroom teaching activities,a special dual clustering processing scheme is formulated,and the application experiment of the algorithm is carried out through the recorded speech data of classroom speech.Through experimental research,it is found that the dual clustering algorithm of GMM-UBM proposed in this study can identify and classify speakers more accurately than simple GMM,and has certain application value.
作者
王娜
刘魏娜
WANG Na;LIU Wei-na(Weinan Vocational&Technological College,Weinan 710000 China;Shaanxi University of Technology,School of Foreign Studies,Hanzhong 723000 China)
出处
《自动化技术与应用》
2024年第6期78-81,共4页
Techniques of Automation and Applications
基金
陕西省教育厅2021年度一般专项科研计划项目(21JK0137)。
关键词
发言者分类算法
课堂教学
会话语料库
聚类
speaker classification algorithm
classroom teaching
discourse corpus
dustering