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. 发明授权

US06917918B2 Method and system for frame alignment and unsupervised adaptation of acoustic models 失效
标题翻译：声学模型的框架对准和无监督适应的方法和系统
公开(公告)号：US06917918B2
公开(公告)日：2005-07-12
申请号：US09746583
申请日：2000-12-22
申请人： William H. Rockenbeck , Milind V. Mahajan , Fileno A. Alleva
发明人： William H. Rockenbeck , Milind V. Mahajan , Fileno A. Alleva
IPC分类号： G10L15/06
CPC分类号： G10L15/065
摘要： An unsupervised adaptation method and apparatus are provided that reduce the storage and time requirements associated with adaptation. Under the invention, utterances are converted into feature vectors, which are decoded to produce a transcript and alignment unit boundaries for the utterance. Individual alignment units and the feature vectors associated with those alignment units are then provided to an alignment function, which aligns the feature vectors with the states of each alignment unit. Because the alignment is performed within alignment unit boundaries, fewer feature vectors are used and the time for alignment is reduced. After alignment, the feature vector dimensions aligned to a state are added to dimension sums that are kept for that state. After all the states in an utterance have had their sums updated, the speech signal and the alignment units are deleted. Once sufficient frames of data have been received to perform adaptive training, the acoustic model is adapted.
摘要翻译：提供了一种无监督的适配方法和装置，其减少与适应相关联的存储和时间要求。根据本发明，话语被转换为特征向量，其被解码以产生用于话语的转录和对准单元边界。然后将单个对准单元和与这些对准单元相关联的特征向量提供给对准功能，其将特征向量与每个对准单元的状态对准。由于在对齐单元边界内进行对齐，所以使用较少的特征向量并减少了对准的时间。对齐后，与状态对齐的特征向量维被添加到为该状态保留的维数和。在话语中所有的状态已经更新了它们的和之后，语音信号和对齐单元被删除。一旦已经接收到足够的数据帧来执行自适应训练，则改变声学模型。

2. 发明授权

US07529671B2 Block synchronous decoding 失效
标题翻译：块同步解码
公开(公告)号：US07529671B2
公开(公告)日：2009-05-05
申请号：US10379244
申请日：2003-03-04
申请人： William H. Rockenbeck , Julian J. Odell
发明人： William H. Rockenbeck , Julian J. Odell
IPC分类号： G10L15/14
CPC分类号： G10L15/142
摘要： A pattern recognition system and method are provided. Aspects of the invention are particularly useful in combination with multi-state Hidden Markov Models. Pattern recognition is effected by processing Hidden Markov Model Blocks. This block-processing allows the processor to perform more operations upon data while such data is in cache memory. By so increasing cache locality, aspects of the invention provide significantly improved pattern recognition speed.
摘要翻译：提供了一种模式识别系统和方法。本发明的各方面与多状态隐马尔科夫模型相结合特别有用。模式识别通过处理隐马尔科夫模型块来实现。该块处理允许处理器在数据处于高速缓冲存储器中时对数据执行更多的操作。通过如此增加缓存局部性，本发明的各方面提供显着改进的模式识别速度。

3. 发明授权

US5752025A Method, computer program product, and system for creating and displaying a categorization table 失效
标题翻译：方法，计算机程序产品和用于创建和显示分类表的系统
公开(公告)号：US5752025A
公开(公告)日：1998-05-12
申请号：US679052
申请日：1996-07-12
申请人： Darren Arthur Shakib , William H. Rockenbeck , Max Loell Benson , Milind Madhukar Joshi
发明人： Darren Arthur Shakib , William H. Rockenbeck , Max Loell Benson , Milind Madhukar Joshi
IPC分类号： G06F17/30
CPC分类号： G06F17/30554 , G06F17/30483 , G06F17/30522 , Y10S707/99943 , Y10S707/99945
摘要： A method, computer program product, and system for creating and displaying a categorization table is provided. A categorization table into a set of data records is created and displayed by first sorting the data records according to data record fields, typically using an index. Next, a header table is created having a number of category columns that will hold values corresponding to a number of index columns. The combinations of values across each header row will be unique and the row will also have a count value indicating the number of data records having the unique combination of category values that can be used to quickly access the first data record in the sorted index. In this manner, all of the data records can be successively accessed knowing the count value and the location of the first data record. Additionally, header rows having no data records associated therewith are used to form the headings and subheadings of the categorization table. Finally, a processor will traverse the header table and cause the categorization table to be displayed on a display means. As part of the traversal, information contained in the header table is used to calculate a fractional position of a user's location with respect to the user's particular view of the categorization table. A plurality of user columns in the categorization table allow separate and distinct views by different users into the same set of data records and using the same header table.
摘要翻译：提供了一种用于创建和显示分类表的方法，计算机程序产品和系统。通过首先根据数据记录字段（通常使用索引）对数据记录进行排序来创建并显示分类表到一组数据记录中。接下来，创建具有多个类别列的标题表，其将保存对应于多个索引列的值。每个标题行之间的值的组合将是唯一的，并且该行还将具有指示具有可用于快速访问排序索引中的第一数据记录的类别值的唯一组合的数据记录的数量的计数值。以这种方式，可以了解第一数据记录的计数值和位置，连续访问所有的数据记录。此外，不使用与其相关联的数据记录的标题行来形成分类表的标题和子标题。最后，处理器将遍历标题表，并使分类表显示在显示装置上。作为遍历的一部分，标题表中包含的信息用于相对于用户的分类表的特定视图计算用户位置的分数位置。分类表中的多个用户列允许不同用户将单独和不同的视图分成同一组数据记录并使用相同的头表。

4. 发明授权

US07016838B2 Method and system for frame alignment and unsupervised adaptation of acoustic models 失效
公开(公告)号：US07016838B2
公开(公告)日：2006-03-21
申请号：US10987529
申请日：2004-11-12
申请人： William H. Rockenbeck , Milind V. Mahajan , Fileno A. Alleva
发明人： William H. Rockenbeck , Milind V. Mahajan , Fileno A. Alleva
IPC分类号： G10L15/06
CPC分类号： G10L15/065
摘要： An unsupervised adaptation method and apparatus are provided that reduce the storage and time requirements associated with adaptation. Under the invention, utterances are converted into feature vectors, which are decoded to produce a transcript and alignment unit boundaries for the utterance. Individual alignment units and the feature vectors associated with those alignment units are then provided to an alignment function, which aligns the feature vectors with the states of each alignment unit. Because the alignment is performed within alignment unit boundaries, fewer feature vectors are used and the time for alignment is reduced. After alignment, the feature vector dimensions aligned to a state are added to dimension sums that are kept for that state. After all the states in an utterance have had their sums updated, the speech signal and the alignment units are deleted. Once sufficient frames of data have been received to perform adaptive training, the acoustic model is adapted.

5. 发明授权

US5915004A Moving a messaging system mailbox 失效
标题翻译：移动邮件系统邮箱
公开(公告)号：US5915004A
公开(公告)日：1999-06-22
申请号：US680217
申请日：1996-07-11
申请人： Ramesh B. Pabbati , Patrick Y. Ng , John H. Howard , William H. Rockenbeck , Max L. Benson
发明人： Ramesh B. Pabbati , Patrick Y. Ng , John H. Howard , William H. Rockenbeck , Max L. Benson
IPC分类号： H04L12/58 , H04M11/00
CPC分类号： H04L51/22 , H04L51/30
摘要： A messaging system user's mailbox is moved using a method in which the mailbox is flagged as being moved and a message arriving for the user during the move is delivered to a holding area to await re-delivery to the mailbox after the move is completed. A combination of non-volatile memory and volatile memory is used to ensure the message's delivery in the event of a power loss or other system failure.
摘要翻译：消息系统用户的邮箱被移动，其中邮箱被标记为被移动，并且在移动期间为用户到达的消息被传送到保持区域，以在移动完成之后等待重新发送到邮箱。使用非易失性存储器和易失性存储器的组合来确保消息在发生电源丢失或其他系统故障时的传送。

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