公开(公告)号：JP2009151318A

公开(公告)日：2009-07-09

申请号：JP2009006033

申请日：2009-01-14

Applicant: Qualcomm Inc ,  クゥアルコム・インコーポレイテッドQualcomm Incorporated

Inventor： CHANG CHIENCHUNG ,  MALAYATH NAREN ,  YAFUSO BYRON YOSHIO

IPC: G10L15/02 ,  G10L15/06 ,  G10L15/20 ,  G10L15/28

CPC classification number: G10L15/065 ,  G10L15/30

Abstract: PROBLEM TO BE SOLVED: To provide a distributed voice recognition system using acoustic feature vector modification.
SOLUTION: The voice recognition system applies speaker-dependent modification functions to acoustic feature vectors prior to voice recognition pattern matching against a speaker-independent acoustic model (238). An adaptation engine (224) matches a set of acoustic feature vectors X with an adaptation model (228) to select a speaker-dependent feature vector modification function f(), which in turn is then applied to X to form a modified set of acoustic feature vectors f(X). Voice recognition is then performed by correlating the modified acoustic feature vectors f(X) with a speaker-independent acoustic model (238).
COPYRIGHT: (C)2009,JPO&INPIT

Abstract translation: 要解决的问题：提供使用声学特征向量修改的分布式语音识别系统。 解决方案：语音识别系统在声音识别模式匹配与扬声器无关的声学模型（238）之前，将声音特征向量应用于扬声器相关的修改功能。 适应引擎（224）将一组声学特征向量X与适配模型（228）匹配，以选择说话者依赖特征向量修改函数f（），然后将其应用于X以形成经修改的声学集合 特征向量f（X）。 然后通过将修改的声学特征向量f（X）与与扬声器无关的声学模型（238）相关联来执行语音识别。 版权所有（C）2009，JPO＆INPIT

公开(公告)号：AT407420T

公开(公告)日：2008-09-15

申请号：AT02702130

申请日：2002-01-30

Applicant: QUALCOMM INC

Inventor： CHANG CHIENCHUNG ,  MALAYATH NAREN ,  YAFUSO BYRON

IPC: G10L15/06 ,  G10L15/28

Abstract: A voice recognition system applies speaker-dependent modification functions to acoustic feature vectors prior to voice recognition pattern matching against a speaker-independent acoustic model. An adaptation engine matches a set of acoustic feature vectors X with an adaptation model to select a speaker-dependent feature vector modification function f( ), which is then applied to X to form a modified set of acoustic feature vectors f(X). Voice recognition is then performed by correlating the modified acoustic feature vectors f(X) with a speaker-independent acoustic model.

公开(公告)号：AU2002235513A1

公开(公告)日：2002-08-28

申请号：AU2002235513

申请日：2002-01-30

Applicant: QUALCOMM INC

Inventor： MALAYATH NAREN ,  YAFUSO BYRON YOSHIO ,  CHANG CHIENCHUNG

IPC: G10L15/06 ,  G10L15/28 ,  G10L15/20

Abstract: A voice recognition system applies speaker-dependent modification functions to acoustic feature vectors prior to voice recognition pattern matching against a speaker-independent acoustic model. An adaptation engine matches a set of acoustic feature vectors X with an adaptation model to select a speaker-dependent feature vector modification function f( ), which is then applied to X to form a modified set of acoustic feature vectors f(X). Voice recognition is then performed by correlating the modified acoustic feature vectors f(X) with a speaker-independent acoustic model.

公开(公告)号：DE60228682D1

公开(公告)日：2008-10-16

申请号：DE60228682

申请日：2002-01-30

Applicant: QUALCOMM INC

Inventor： CHANG CHIENCHUNG ,  MALAYATH NAREN ,  YAFUSO BYRON YOSHIO

IPC: G10L15/06 ,  G10L15/28

Abstract: A voice recognition system applies speaker-dependent modification functions to acoustic feature vectors prior to voice recognition pattern matching against a speaker-independent acoustic model. An adaptation engine matches a set of acoustic feature vectors X with an adaptation model to select a speaker-dependent feature vector modification function f( ), which is then applied to X to form a modified set of acoustic feature vectors f(X). Voice recognition is then performed by correlating the modified acoustic feature vectors f(X) with a speaker-independent acoustic model.

公开(公告)号：BR0206836A

公开(公告)日：2006-01-17

申请号：BR0206836

申请日：2002-01-30

Applicant: QUALCOMM INC

Inventor： CHANG CHIENCHUNG ,  MALAYATH NAREN ,  YAFUSO BYRON YOSHIO

IPC: G10L15/06 ,  G10L15/28

Abstract: A voice recognition system applies speaker-dependent modification functions to acoustic feature vectors prior to voice recognition pattern matching against a speaker-independent acoustic model. An adaptation engine matches a set of acoustic feature vectors X with an adaptation model to select a speaker-dependent feature vector modification function f( ), which is then applied to X to form a modified set of acoustic feature vectors f(X). Voice recognition is then performed by correlating the modified acoustic feature vectors f(X) with a speaker-independent acoustic model.

公开(公告)号：HK1062738A1

公开(公告)日：2004-11-19

申请号：HK04105572

申请日：2004-07-28

Applicant: QUALCOMM INC

Inventor： CHANG CHIENCHUNG ,  MALAYATH NAREN ,  YAFUSO BYRON YOSHIO

IPC: G10L20060101 ,  G10L15/06 ,  G10L15/28

Abstract: A voice recognition system applies speaker-dependent modification functions to acoustic feature vectors prior to voice recognition pattern matching against a speaker-independent acoustic model. An adaptation engine matches a set of acoustic feature vectors X with an adaptation model to select a speaker-dependent feature vector modification function f( ), which is then applied to X to form a modified set of acoustic feature vectors f(X). Voice recognition is then performed by correlating the modified acoustic feature vectors f(X) with a speaker-independent acoustic model.

公开(公告)号：AU2003238294A1

公开(公告)日：2003-12-31

申请号：AU2003238294

申请日：2003-06-18

Applicant: QUALCOMM INC

Inventor： MALAYATH NAREN ,  CHANG CHIENCHUNG ,  JALIL SUHAIL

IPC: G06T9/00 ,  H03M7/36 ,  H04N19/51 ,  H04N7/26

Abstract: Video encoding techniques are described. In one example, a video encoding technique includes identifying a pixel location associated with a video block in a search space based on motion vectors associated with a set of video blocks within a video frame to be encoded, wherein the video blocks in the set are spatially located at defined locations relative to a current video block of the video frame to be encoded. A motion estimation routine can then be initialized for the current video block at the identified pixel location. By identifying a pixel location associated with a video block in a search space based on motion vectors associated with a set of video blocks within a video frame, the phenomenon of spatial redundancy can be more readily exploited to accelerate and improve the encoding process.

公开(公告)号：WO02065453A3

公开(公告)日：2002-10-24

申请号：PCT/US0203014

申请日：2002-01-30

Applicant: QUALCOMM INC

Inventor： CHANG CHIENCHUNG ,  MALAYATH NAREN ,  YAFUSO BYRON YOSHIO

IPC: G10L15/06 ,  G10L15/28 ,  G10L15/20

CPC classification number: G10L15/065 ,  G10L15/30

Abstract: A voice recognition system applies speaker-dependent modification functions to acoustic feature vectors prior to voice recognition pattern matching against a speaker-independent acoustic model (238). An adaptation engine (224) matches a set of acoustic feature vectors X with an adaptation model (228) to select a speaker-dependent feature vector modification function f(), which is then applied to X to form a modified set of acoustic feature vectors f(X). Voice recognition is then performed by correlating the modified acoustic feature vectors f(X) with a speaker-independent acoustic model (238).

Abstract translation: 在语音识别模式匹配与说话者无关的声学模型（238）之前，语音识别系统将与说话者有关的修改函数应用于声学特征向量。 自适应引擎（224）将一组声学特征向量X与自适应模型（228）进行匹配以选择随讲话者而变的特征向量修改函数f（），然后将其应用于X以形成修改的声学特征向量集合 F（X）。 然后通过将修改的声学特征向量f（X）与不依赖于说话者的声学模型（238）进行相关来执行语音识别。

公开(公告)号：WO03107680A2

公开(公告)日：2003-12-24

申请号：PCT/US0319400

申请日：2003-06-18

Applicant: QUALCOMM INC

Inventor： MALAYATH NAREN ,  CHANG CHIENCHUNG ,  JALIL SUHAIL

IPC: G06T9/00 ,  H03M7/36 ,  H04N19/51 ,  H04N7/26

CPC classification number: H04N19/127 ,  H04N19/51 ,  H04N19/557 ,  H04N19/56 ,  H04N19/57

Abstract: Video encoding techniques are described. In one example, a video encoding technique includesidentifying a pixel location associated with a video block in a search space based on motion vectors associated with a set of video blocks within a video frame to be encoded, wherein the video blocks in the set are spatially located at defined locations relative to a current video block of the video frame to be encoded. A motion estimation routine can then be initialized for the current video block at the identified pixel location. By identifying a pixel location associated with a video block in a search space based on motion vectors associated with a set of video blocks within a video frame, the phenomenon of spatial redundancy can be more readily exploited to accelerate and improve the encoding process.

Abstract translation: 描述视频编码技术。 在一个示例中，视频编码技术包括基于与要编码的视频帧内的一组视频块相关联的运动矢量来识别与搜索空间中的视频块相关联的像素位置，其中该组中的视频块位于空间上 在相对于要编码的视频帧的当前视频块的定义位置处。 然后可以针对所识别的像素位置处的当前视频块初始化运动估计例程。 通过基于与视频帧内的一组视频块相关联的运动矢量来识别与搜索空间中的视频块相关联的像素位置，可以更容易地利用空间冗余现象来加速和改进编码处理。