专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明申请

WO2001048737A3 SPEECH RECOGNIZER WITH A LEXICAL TREE BASED N-GRAM LANGUAGE MODEL 审中-公开
公开(公告)号：WO2001048737A3
公开(公告)日：2001-07-05
申请号：PCT/CN1999/000217
申请日：1999-12-23
申请人： INTEL CORPORATION , LIN, Zhiwei , YAN, Yonghong , ZHAO, Qingwei , YUAN, Baosheng
发明人： LIN, Zhiwei , YAN, Yonghong , ZHAO, Qingwei , YUAN, Baosheng
IPC分类号： G10L15/06
摘要： In some embodiments, the invention includes a method comprising creating a lexical tree and identifying beginning phonemes in the lexical tree. The method of these embodiments further includes estimating probabilities of words in the lexical tree having particular ones of the beginning phonemes and storing at least some of the estimated probabilities, wherein backoff weights are not stored with the estimated probabilities. The estimated probabilities may be stored in a lookup table. In other embodiment, the invention includes a method of receiving phonemes and identifying them on a lexical tree. The method of these embodiments also includes estimating probabilities of words that include the phonemes through use of estimated probabilities retrieved from storage, wherein the retrieve probabilities do not include backoff weights stored with the estimated probabilities. Again, the estimated probabilities may be stored in a lookup table. The estimated probabilities may be used in establishing a pruning threshold. The methods may be implemented by instructions on a computer readable medium.

2. 发明申请

WO2002103675A1 CLIENT-SERVER BASED DISTRIBUTED SPEECH RECOGNITION SYSTEM ARCHITECTURE 审中-公开
标题翻译：基于客户端的分布式语音识别系统架构
公开(公告)号：WO2002103675A1
公开(公告)日：2002-12-27
申请号：PCT/CN2001/001030
申请日：2001-06-19
申请人： INTEL CORPORATION , INTEL CHINA LTD. , ZHAO, Qingwei , ZHANG, Xiangdong , YAN, Yonghong , YUAN, Baosheng
发明人： ZHAO, Qingwei , ZHANG, Xiangdong , YAN, Yonghong , YUAN, Baosheng
IPC分类号： G10L15/00
CPC分类号： G10L15/30
摘要： In general, the new client-server based Distributed Speech Recognition (DSR) system provides an effective method of recognizing speech made by a human at a client device and transmitted to a remote server over a network. The system distributes the speech recognition process between the client and the server so that a speaker-dependent language model may be utilized yielding higher accuracy as compared to the tradition DSR systems. Accordingly, the client device is configured to generate a phonetic word graph by performing acoustic recognition using an acoustic model that is trained by the same end-user whose speech is to be recognized. The resulting phonetic word graph is transmitted to the server which will handle the language processing and generate a recognized word sequence. When compared to a design that uses the traditional DSR, the new DSR method and system produces a word error rate that is at least 2-3 times lower, resulting in a higher accuracy recognition system.
摘要翻译：通常，基于新的基于客户端 - 服务器的分布式语音识别（DSR）系统提供了一种有效的方法来识别由客户端设备发送的人的语音，并通过网络发送到远程服务器。该系统在客户端和服务器之间分配语音识别过程，使得与传统DSR系统相比，可以利用与说话者相关的语言模型产生更高的精度。因此，客户端设备被配置为通过使用由要被识别其语音的相同终端用户训练的声学模型进行声学识别来生成语音字图。所产生的语音字图被发送到将处理语言处理并产生识别的字序列的服务器。与使用传统DSR的设计相比，新的DSR方法和系统产生的字错误率至少低2-3倍，导致更高的精度识别系统。

3. 发明申请

WO2002091357A1 METHOD, APPARATUS, AND SYSTEM FOR BUILDING CONTEXT DEPENDENT MODELS FOR A LARGE VOCABULARY CONTINUOUS SPEECH RECOGNITION (LVCSR) SYSTEM 审中-公开
标题翻译：用于建立大型语音连续语音识别（LVCSR）系统的上下文相关模型的方法，装置和系统
公开(公告)号：WO2002091357A1
公开(公告)日：2002-11-14
申请号：PCT/CN2001/000684
申请日：2001-05-08
申请人： INTEL CORPORATION , INTEL CHINA LTD. , LIU, Xiaoxing , YUAN, Baosheng , YAN, Yonghong
发明人： LIU, Xiaoxing , YUAN, Baosheng , YAN, Yonghong
IPC分类号： G10L15/06
CPC分类号： G10L15/187 , G10L15/1815
摘要： According to one aspect of the invention, a method is provided in which a set of multiple mixture monophone models is created and trained to generate a set of multiple mixture context dependent models. A set of single mixture triphone models is created and trained to generate a set of context dependent models. Corresponding states of the triphone models are clustered to obtain a set of tied states based on a decision tree clustering process. Parameters of the context dependent models are estimated using a data dependent maximum a posteriori (MAP) adaptation method in which parameters of the tied states of the context dependent models are derived by adapting corresponding parameters of the context independent models using the training data associated with the respective tied states.
摘要翻译：根据本发明的一个方面，提供了一种方法，其中创建并训练一组多个混合单声道模型以生成一组多个混合上下文相关模型。创建和训练了一组单一混合三音模型，以生成一组上下文相关模型。将三通电话模型的对应状态聚类成基于决策树聚类过程获得一组绑定状态。使用依赖于数据的最大后验（MAP）适配方法来估计上下文相关模型的参数，其中，通过使用与上下文相关模型相关联的训练数据来调整上下文无关模型的相应参数，从而导出上下文相关模型的绑定状态的参数各自的绑定状态。

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式