QQ客服 官方微博 反馈留言

	Structure constrained distributions matching using quadratic programming and its application to pronunciation evaluation
	Yu Qiao; Masayuki Suzuki; Nobuaki Minematsu; Keikichi Hirose
	2011
会议名称	1st Asian Conference on Pattern Recognition (ACPR)
会议地点	Beijing, PEOPLES R CHINA
英文摘要	We proposed a structural representation of speech that is robust to speaker difference due to its transformation-invariant property in previous works, where we compared two speech structures by calculating the distance between two structural vectors, each composed of the lengths of a structure's edges. However, this distance can not yield matching scores directly related to individual events (nodes) of the two structures. In spite of comparing structural vectors directly, this paper takes structures as constraints for optimal pattern matching. We derive the formulas of objective functions and constraint functions for optimization. Under assumptions of Gaussian and shared covariance matrices, we show that this optimal problem can be reduced to a quadratically constrained quadraticprogramming problem. To relieve the too strong invariance problem, we use a subspace decomposition method and perform the optimization in each subspace. We evaluate the proposed method on a task to assess the goodness of students' English pronunciation. Experimental results show that the proposed method achieves higher correlations with teachers' manuals cores than compared methods.
收录类别	EI
语种	英语
内容类型	会议论文
源URL	[http://ir.siat.ac.cn:8080/handle/172644/3262]
专题	深圳先进技术研究院_集成所
作者单位	2011
推荐引用方式 GB/T 7714	Yu Qiao,Masayuki Suzuki,Nobuaki Minematsu,et al. Structure constrained distributions matching using quadratic programming and its application to pronunciation evaluation[C]. 见:1st Asian Conference on Pattern Recognition (ACPR). Beijing, PEOPLES R CHINA.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。