公开(公告)号：WO2008002881A2

公开(公告)日：2008-01-03

申请号：PCT/US2007072039

申请日：2007-06-25

Applicant: QUALCOMM INC ,  GARUDADRI HARINATH ,  REZNIK YURIY

Inventor： GARUDADRI HARINATH ,  REZNIK YURIY

IPC: G06F17/14

CPC classification number: G06F17/147 ,  H04N19/45 ,  H04N19/60 ,  H04N19/61

Abstract: Techniques are described to reduce rounding errors during computation of discrete cosine transform using fixed-point calculations. According to these techniques, a discrete cosine transform a matrix of scaled coefficients is calculated by multiplying coefficients in a matrix of coefficients by scale factors. Next, a midpoint bias value and a supplemental bias value are added to a DC coefficient of the matrix of scaled coefficients. Next, an inverse discrete cosine transform is applied to the resulting matrix of scaled coefficients. Values in the resulting matrix are then right-shifted in order to derive a matrix of pixel component values. As described herein, the addition of the supplemental bias value to the DC coefficient reduces rounding errors attributable to this right-shifting. As a result, a final version of a digital media file decompressed using these techniques may more closely resemble an original version of a digital media file.

Abstract translation: 描述了使用定点计算在离散余弦变换计算期间减少舍入误差的技术。 根据这些技术，通过用系数矩阵乘以比例因子来计算缩放系数矩阵的离散余弦变换。 接下来，将中点偏置值和补充偏置值加到缩放系数矩阵的DC系数中。 接下来，将逆离散余弦变换应用于所得到的缩放系数矩阵。 然后将所得矩阵中的值右移，以便导出像素分量值的矩阵。 如这里所述，补充偏置值加到DC系数可以减少归因于该右移的舍入误差。 因此，使用这些技术解压缩的数字媒体文件的最终版本可能更接近于数字媒体文件的原始版本。

公开(公告)号：WO2007047478A2

公开(公告)日：2007-04-26

申请号：PCT/US2006040165

申请日：2006-10-12

Applicant: QUALCOMM INC ,  REZNIK YURIY ,  CHUNG HYUKJUNE ,  GARUDADRI HARINATH ,  SRINIVASAMURTHY NAVEEN D ,  SAGETONG PHOOM

Inventor： REZNIK YURIY ,  CHUNG HYUKJUNE ,  GARUDADRI HARINATH ,  SRINIVASAMURTHY NAVEEN D ,  SAGETONG PHOOM

CPC classification number: G06F17/147 ,  H03H17/0225 ,  H04N19/42 ,  H04N19/60 ,  H04N19/61 ,  H04N19/70

Abstract: Techniques for efficiently performing computation for signal and data processing are described. For multiplication-free processing, a series of intermediate values is generated based on an input value for data to be processed. At least one intermediate value in the series is generated based on at least one other intermediate value in the series. One intermediate value in the series is provided as an output value for a multiplication of the input value with a constant value. The constant value may be an integer constant, a rational constant, or an irrational constant. An irrational constant may be approximated with a rational dyadic constant having an integer numerator and a denominator that is a power of twos. The multiplication-free processing may be used for various transforms (e.g., DCT and IDCT), filters, and other types of signal and data processing.

Abstract translation: 描述了用于有效执行信号和数据处理计算的技术。 对于无需乘法处理，根据要处理的数据的输入值生成一系列中间值。 基于该系列中的至少一个其它中间值，生成该系列中的至少一个中间值。 该系列中的一个中间值被提供为用于将输入值与常数值相乘的输出值。 常数值可以是整数常数，有理常数或非理性常数。 非理性常数可以用具有整数分子和二分之一的分母的有理二元常数近似。 无乘法处理可以用于各种变换（例如，DCT和IDCT），滤波器以及其它类型的信号和数据处理。

公开(公告)号：WO2005114943A3

公开(公告)日：2006-01-19

申请号：PCT/US2005016837

申请日：2005-05-13

Applicant: QUALCOMM INC ,  GARUDADRI HARINATH ,  SAGETONG PHOOM ,  NANDA SANJIV ,  LUNDBY STEIN A

Inventor： GARUDADRI HARINATH ,  SAGETONG PHOOM ,  NANDA SANJIV ,  LUNDBY STEIN A

IPC: H04B7/00 ,  H04B7/216 ,  H04L12/28 ,  H04L12/56 ,  H04L12/66 ,  H04L29/06 ,  H04N7/26 ,  H04W28/06 ,  H04W72/12 ,  H04W84/04 ,  H04W88/18 ,  H04Q7/30

CPC classification number: H04W28/06 ,  H04L29/06 ,  H04L29/06027 ,  H04L47/36 ,  H04L65/601 ,  H04L65/604 ,  H04L65/80 ,  H04L69/04 ,  H04L69/161 ,  H04L69/166 ,  H04L69/22 ,  H04N19/102 ,  H04N19/115 ,  H04N19/124 ,  H04N19/152 ,  H04N19/164 ,  H04N19/174 ,  H04N19/61 ,  H04N21/2381 ,  H04N21/41407 ,  H04N21/44004 ,  H04N21/4788 ,  H04N21/6131 ,  H04N21/6181 ,  H04N21/6437 ,  H04N21/64707 ,  H04W4/06 ,  H04W28/065 ,  H04W72/044 ,  H04W72/1263 ,  H04W80/00 ,  H04W84/04 ,  H04W88/02 ,  H04W88/181

Abstract: Methods and apparatus are described for transmitting information units over a plurality of constant bit rate communication channel. The techniques include encoding the information units, thereby creating a plurality of data packets. The encoding is constrained such that the data packet sizes match physical layer packet sizes of the communication channel. The information units may include a variable bit rate data stream, multimedia data, video data, and audio data. The communication channels include CMDA channels, WCDMA, GSM channels, GPRS channels, and EDGE channels.

Abstract translation: 描述了用于通过多个恒定比特率通信信道发送信息单元的方法和装置。 这些技术包括对信息单元进行编码，从而创建多个数据分组。 编码被约束，使得数据分组大小与通信信道的物理层分组大小相匹配。 信息单元可以包括可变比特率数据流，多媒体数据，视频数据和音频数据。 通信信道包括CMDA信道，WCDMA，GSM信道，GPRS信道和EDGE信道。

公开(公告)号：WO02080142A3

公开(公告)日：2003-03-13

申请号：PCT/US0208727

申请日：2002-03-22

Applicant: QUALCOMM INC

Inventor： MALAYATH NARENDRANATH ,  DEJACO ANDREW P ,  CHANG CHIENCHUNG ,  JALIL SUHAIL ,  BI NING ,  GARUDADRI HARINATH

IPC: G10L15/06 ,  G10L15/10 ,  G10L15/12 ,  G10L15/14 ,  G10L15/16 ,  G10L17/04

CPC classification number: G10L15/142 ,  G10L15/07 ,  G10L15/10 ,  G10L15/12 ,  G10L15/144 ,  G10L15/32

Abstract: A voice recognition (VR) system is disclosed that utilizes a combination of speaker independent (SI) (230 and 232) and speaker dependent (SD) (234) acoustic models. At least one SI acoustic model (230 and 232) is used in combination with at least one SD acoustic model (234) to provide a level of speech recognition performance that at least equals that of a purely SI acoustic model. The disclosed hybrid SI/SD VR system continually uses unsupervised training to update the acoustic templates in the one ore more SD acoustic models (234). The hybrid VR system then uses the updated SD acoustic models (234) in combination with the at least one SI acoustic model (230 and 232) to provide improved VR performance during VR testing.

Abstract translation: 公开了一种利用独立于扬声器（SI）（230和232）和扬声器依赖（SD）（234）声学模型的组合的语音识别（VR）系统。 至少一个SI声学模型（230和232）与至少一个SD声学模型（234）组合使用，以提供至少等于纯SI声学模型的语音识别性能的水平。 所公开的混合SI / SD VR系统连续地使用无监督的训练来更新一个或多个SD声学模型中的声学模板（234）。 混合VR系统然后使用更新的SD声学模型（234）与至少一个SI声学模型（230和232）组合，以在VR测试期间提供改进的VR性能。

公开(公告)号：WO02061727A3

公开(公告)日：2003-02-27

申请号：PCT/US0202625

申请日：2002-01-29

Applicant: QUALCOMM INC

Inventor： GARUDADRI HARINATH ,  HERMANSKY HYNEK ,  BURGET LUKAS ,  JAIN PRATIBHA ,  KAJAREKAR SACHIN ,  SIVADAS SUNIL ,  DUPONT STEPHANE N ,  ORTUZAR MARIA CARMEN BENITEZ ,  MORGAN NELSON H

IPC: G10L15/02 ,  G10L15/28 ,  G10L11/06

CPC classification number: G10L15/30 ,  G10L15/02

Abstract: A system and method for extracting acoustic features and speech activity on a device and transmitting them in a distributed voice recognition system. The distributed voice recognition system includes a local VR engine in a subscriber unit (102) and a server VR engine in a server (160). The local VR engine comprises a feature extraction (FE) module (104) that extracts features from a speech signal, and a voice activity detection module (VAD) (106) that detects voice activity within a speech signal. The voice activity signal and the features are downsampled before they are transmitted from the local engine to the server engine. The system includes filters, framing and windowing modules, power spectrum analyzers, a neural network, a nonlinear element, and other components to selectively provide an advanced front end vector including predetermined portions of the voice activity detection indication and extracted features from the subscriber unit (104) to the server (160). The indication of detected voice activity is transmitted ahead of the extracted features in order to avoid long recognition delays. The system also includes a module to generate additional feature vectors on the server from the received features using a feed-forward multilayer perception (MLP) and providing the same to the speech server (160).

Abstract translation: 一种用于在设备上提取声学特征和语音活动并在分布式语音识别系统中传送它们的系统和方法。 分布式语音识别系统包括用户单元（102）中的本地VR引擎和服务器（160）中的服务器VR引擎。 本地VR引擎包括从语音信号中提取特征的特征提取（FE）模块（104）和检测语音信号内的语音活动的语音活动检测模块（VAD）（106）。 语音活动信号和特征在从本地引擎传输到服务器引擎之前被下采样。 该系统包括滤波器，成帧和开窗模块，功率谱分析仪，神经网络，非线性元件和其他组件，以选择性地提供包括来自用户单元的语音活动检测指示和提取特征的预定部分的高级前端矢量 104）发送到服务器（160）。 检测到的语音活动的指示在提取的特征之前传输，以避免长的识别延迟。 该系统还包括一个模块，用于使用前馈多层感知（MLP）从所接收的特征在服务器上产生附加的特征向量，并将其提供给语音服务器（160）。