Speech models generated using competitive training, asymmetric training, and data boosting
作者： Xiaodong He , Jian Wu
DOI:
关键词:
摘要: Speech models are trained using one or more of three different training systems. They include competitive which reduces a distance between recognized result and true result, data boosting divides weights data, asymmetric trains model components differently.
google.de 参考文章(17)
Jean-claude Junqua, Christophe Cerisara, Luca Rigazio, Robert Boman, Method for additive and convolutional noise adaptation in automatic speech recognition using transformed matrices ,(2000)
Ying Jia, Yonghong Yan, Baosheng Yuan, Method, apparatus, and system for bottom-up tone integration to chinese continuous speech recognition system ,(2000)
John Grothendieck, Allen Louis Gorin, Jeremy Huntley Greet Wright, Apparatus and method for analysis of language model changes ,(2006)
Soichi Toyama, Apparatus and method for speech recognition ,(2003)
Arul A. Menezes, Eric D. Brill, Method and apparatus for unsupervised training of natural language processing units ,(2001)
Oren Glickman, Christopher Frank Joerg, Method for aligning text with audio signals ,(1998)
Makoto Shozakai, Goshu Nagino, Data process unit and data process unit control program ,(2004)
Aleš Padrta, Vlasta Radová, On the background model construction for speaker verification using GMM text speech and dialogue. pp. 425- 432 ,(2004) , 10.1007/978-3-540-30120-2_54
Robert L. Rounthwaite, David E. Heckerman, John L. Manferdelli, Fileno A. Alleva, Daniel Rosen, Mei-Yuh Hwang, Yoram Yaacovi, Methods and apparatus for performing speech recognition using acoustic models which are improved through an interactive process ,(2000)