Feature space generalized variable parameter HMMs for noise robust recognition | |
Yang Li; Xunying Liu; Lan Wang | |
2013 | |
会议名称 | 14th Annual Conference of the International Speech Communication Association, INTERSPEECH 2013 |
会议地点 | Lyon, France |
英文摘要 | Handling variable ambient noise is a challenging task for automatic speech recognition (ASR) systems. To address this issue, multi-style training using speech data collected in diverse noise environments, noise adaptive training or uncertainty decoding techniques can be used. An alternative approach is to explicitly approximate the continuous trajectory of Gaussian component or model space linear transform parameters against the varying noise, for example, using generalized variable parameter HMMs (GVP-HMM). In order to reduce the computational cost of conventional GVP-HMMs when model parameter update against the varying noise condition is required, this paper investigates a novel and more efficient extension of GVPHMMs that can also model the trajectories of feature space linear transforms. Significant error rate reductions of 9.3% and 18.5% relative were obtained over the multi-style training baseline system on Aurora 2 and a medium vocabulary Mandarin Chinese speech recognition task respectively. |
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/172644/4505] |
专题 | 深圳先进技术研究院_集成所 |
作者单位 | 2013 |
推荐引用方式 GB/T 7714 | Yang Li,Xunying Liu,Lan Wang. Feature space generalized variable parameter HMMs for noise robust recognition[C]. 见:14th Annual Conference of the International Speech Communication Association, INTERSPEECH 2013. Lyon, France. |
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。
修改评论