知搜-全球专利检索

  1. Login
  2. Sign Up

Search Results

31 records (took 0.035 seconds)

    Noise adaptive training for speech recognition
    1.
    发明授权
    Noise adaptive training for speech recognition 有权
    语音识别噪声适应训练

    公开(公告)号：US09009039B2

    公开(公告)日：2015-04-14

    申请号：US12483262

    申请日：2009-06-12

    Applicant: Michael Lewis Seltzer ,  James Garnet Droppo ,  Ozlem Kalinli ,  Alejandro Acero

    Inventor： Michael Lewis Seltzer ,  James Garnet Droppo ,  Ozlem Kalinli ,  Alejandro Acero

    IPC: G10L15/20 ,  G10L15/06 ,  G10L15/14

    CPC classification number: G10L15/063 ,  G10L15/144 ,  G10L15/20

    Abstract: Technologies are described herein for noise adaptive training to achieve robust automatic speech recognition. Through the use of these technologies, a noise adaptive training (NAT) approach may use both clean and corrupted speech for training. The NAT approach may normalize the environmental distortion as part of the model training. A set of underlying “pseudo-clean” model parameters may be estimated directly. This may be done without point estimation of clean speech features as an intermediate step. The pseudo-clean model parameters learned from the NAT technique may be used with a Vector Taylor Series (VTS) adaptation. Such adaptation may support decoding noisy utterances during the operating phase of a automatic voice recognition system.

    Abstract translation: 这里描述了用于噪声自适应训练以实现鲁棒自动语音识别的技术。 通过使用这些技术，噪声自适应训练（NAT）方法可以使用干净和损坏的语音进行训练。 NAT方法可以将环境变形归一化，作为模型训练的一部分。 可以直接估计一组潜在的“伪清理”模型参数。 这可以在没有将干净的语音特征的点估计作为中间步骤的情况下完成。 从NAT技术学习的伪清理模型参数可以与矢量泰勒级数（VTS）适配一起使用。 这种适配可以支持在自动语音识别系统的操作阶段期间解码噪声话语。

    Keystroke sound suppression
    3.
    发明授权
    Keystroke sound suppression 有权
    击键声音抑制

    公开(公告)号：US08213635B2

    公开(公告)日：2012-07-03

    申请号：US12328789

    申请日：2008-12-05

    Applicant: Qin Li ,  Michael Lewis Seltzer ,  Chao He

    Inventor： Qin Li ,  Michael Lewis Seltzer ,  Chao He

    IPC: H04B15/00 ,  G10L11/06

    CPC classification number: G10L21/0208 ,  G10L21/0232

    Abstract: An audio signal is received that might include keyboard noise and speech. The audio signal is digitized and transformed from a time domain to a frequency domain. The transformed audio is analyzed to determine whether there is likelihood that keystroke noise is present. If it is determined there is high likelihood that the audio signal contains keystroke noise, a determination is made as to whether a keyboard event occurred around the time of the likely keystroke noise. If it is determined that a keyboard event occurred around the time of the likely keystroke noise, a determination is made as to whether speech is present in the audio signal around the time of the likely keystroke noise. If no speech is present, the keystroke noise is suppressed in the audio signal. If speech is detected in the audio signal or if the keystroke noise abates, the suppression gain is removed from the audio signal.

    Abstract translation: 接收到可能包括键盘噪音和语音的音频信号。 音频信号被数字化并从时域变换到频域。 分析变换的音频以确定是否存在按键噪声的可能性。 如果确定音频信号包含击键噪声的可能性很高，则确定在可能的击键噪声的时间周围是否发生键盘事件。 如果确定在可能的击键噪声的时间周围发生键盘事件，则确定在可能的击键噪声的时间周围的音频信号中是否存在语音。 如果不存在语音，则在音频信号中抑制击键噪声。 如果在音频信号中检测到语音，或者如果击键噪声消失，则从音频信号中去除抑制增益。

    Acoustic Model Adaptation Using Splines
    4.
    发明申请
    Acoustic Model Adaptation Using Splines 有权
    使用样条的声学模型适应

    公开(公告)号：US20110238416A1

    公开(公告)日：2011-09-29

    申请号：US12730270

    申请日：2010-03-24

    Applicant: Michael Lewis Seltzer ,  Kaustubh Prakash Kalgaonkar ,  Alejandro Acero

    Inventor： Michael Lewis Seltzer ,  Kaustubh Prakash Kalgaonkar ,  Alejandro Acero

    IPC: G10L15/20

    CPC classification number: G10L15/20

    Abstract: Described is a technology by which a speech recognizer is adapted to perform in noisy environments using linear spline interpolation to approximate the nonlinear relationship between clean speech, noise, and noisy speech. Linear spline parameters that minimize the error the between predicted noisy features and actual noisy features are learned from training data, along with variance data that reflect regression errors. Also described is compensating for linear channel distortion and updating noise and channel parameters during speech recognition decoding.

    Abstract translation: 描述了一种技术，通过该技术，语音识别器适于在噪声环境中使用线性样条插值来执行，以近似清洁语音，噪声和噪声语音之间的非线性关系。 从训练数据以及反映回归误差的方差数据中可以看出，将预测噪声特征与实际噪声特征之间的误差最小化的线性样条参数。 还描述了在语音识别解码期间补偿线性信道失真和更新噪声和信道参数。

    ADAPTIVE BEAMFORMER USING A LOG DOMAIN OPTIMIZATION CRITERION
    5.
    发明申请
    ADAPTIVE BEAMFORMER USING A LOG DOMAIN OPTIMIZATION CRITERION 有权
    自适应光束使用日志域优化标准

    公开(公告)号：US20100177908A1

    公开(公告)日：2010-07-15

    申请号：US12353986

    申请日：2009-01-15

    Applicant: Michael Lewis Seltzer ,  Ivan Jelev Tashev

    Inventor： Michael Lewis Seltzer ,  Ivan Jelev Tashev

    IPC: H04R3/00

    CPC classification number: H04R1/406 ,  H04R3/005

    Abstract: Described is a audio signal processing technology in which an adaptive beamformer processes input signals from microphones based on an estimate received from a pre-filter. The adaptive beamformer may compute its parameters (e.g., weights) for each frame based on the estimate, via a magnitude-domain objective function or log-magnitude-domain objective function. The pre-filter may include a time invariant beamformer and/or a non-linear spatial filter, and/or may include a spectral filter. The computed parameters may be adjusted based on a constraint, which may be selectively applied only at desired times.

    Abstract translation: 描述了一种音频信号处理技术，其中自适应波束形成器基于从预滤波器接收到的估计来处理来自麦克风的输入信号。 自适应波束形成器可以经由幅度域目标函数或对数幅度域目标函数基于估计来计算其每个帧的参数（例如权重）。 预滤波器可以包括时不变波束形成器和/或非线性空间滤波器，和/或可以包括频谱滤波器。 可以基于约束来调整计算的参数，该约束可以仅在期望的时间选择性地应用。

    ROUTE REWARD AUGMENTATION
    6.
    发明申请
    ROUTE REWARD AUGMENTATION 审中-公开
    路由奖励计划

    公开(公告)号：US20090210302A1

    公开(公告)日：2009-08-20

    申请号：US12033652

    申请日：2008-02-19

    Applicant: Ivan J. Tashev ,  Jeffrey D. Couckuyt ,  John C. Krumm ,  Ruston Panabaker ,  Michael Lewis Seltzer ,  Neil W. Black

    Inventor： Ivan J. Tashev ,  Jeffrey D. Couckuyt ,  John C. Krumm ,  Ruston Panabaker ,  Michael Lewis Seltzer ,  Neil W. Black

    IPC: G06Q30/00 ,  G01C21/00

    CPC classification number: G06Q30/02 ,  G06Q10/047 ,  G06Q30/0273

    Abstract: Different incentives can be provided to a user for the user to take different routes. Information related to a user's specific situation can be gathered and a reward can be associated with a route. A user can input an intended destination and different routes can be taken, where the routes have various characteristics. The route can be analyzed and a determination can be made as to what routes can be associated with a user taking a particular route. Commonly, different companies can supply rewards such that the user travels past their advertisement or makes a stop at their store.

    Abstract translation: 可以向用户提供不同的激励以使用户采取不同的路由。 可以收集与用户特定情况相关的信息，并且可以将奖励与路线相关联。 用户可以输入预期的目的地，并且可以采用不同的路由，其中​​路由具有各种特征。 可以分析路线，并且可以确定可以与采取特定路线的用户相关联的路线的确定。 通常，不同的公司可以提供奖励，使用户经过他们的广告或停止在他们的商店。

    AD HOC AND OPPORTUNISTIC TRANSPORTATION SERVICES
    7.
    发明申请
    AD HOC AND OPPORTUNISTIC TRANSPORTATION SERVICES 审中-公开
    国际和机场运输服务

    公开(公告)号：US20090210276A1

    公开(公告)日：2009-08-20

    申请号：US12033713

    申请日：2008-02-19

    Applicant: John C. Krumm ,  Eric J. Horvitz ,  Ruston Panabaker ,  Michael Lewis Seltzer ,  Neil W. Black ,  Jeffrey D. Couckuyt ,  Ivan J. Tashev

    Inventor： John C. Krumm ,  Eric J. Horvitz ,  Ruston Panabaker ,  Michael Lewis Seltzer ,  Neil W. Black ,  Jeffrey D. Couckuyt ,  Ivan J. Tashev

    IPC: G06Q10/00

    CPC classification number: G06Q10/06 ,  G06Q10/025 ,  G06Q30/0204 ,  G06Q30/0273 ,  G06Q50/30

    Abstract: Individuals can be organized into travel groups in plans constructed in advance or in real-time to save resources and travel in an efficient manner. Multi-segment trips between starting points and destinations can be constructed with several vehicles, including private, public, and commercial transportation assets. Numerous requests for real-time or planned recurrent commutes by a population of users can be considered in a larger analysis that seeks to optimize one or more attributes such as vehicle usage and/or greenhouse emissions. Data concerning multiple related individuals can be gathered and analyzed—based upon the analysis, a determination can be made if it is logical to group individuals together such that they physically travel together. A transportation asset provider and/or individuals can be offered a reward to become part of the travel group and/or to perform specific tasks related to the travel group, such as using a vehicle with a certain fuel type.

    Abstract translation: 个人可以组织成旅行团体，提前或实时制定计划，以节省资源和有效率地旅行。 起点和目的地之间的多段旅行可以用多种车辆构建，包括私人，公共和商业运输资产。 可以在更大的分析中考虑许多用户的实时或计划的经常性通勤请求，以寻求优化一个或多个属性，如车辆使用和/或温室气体排放。 基于分析可以收集和分析关于多个相关个人的数据，如果将个体组合在一起，使得它们在一起物理地旅行是合乎逻辑的，则可以确定。 可以向交通运输资产提供者和/或个人提供奖励以成为旅行团体的一部分和/或执行与旅行团体相关的特定任务，例如使用具有某种燃料类型的车辆。

    AUTOMATIC SPLICES FOR TARGETED ADVERTISEMENTS
    9.
    发明申请
    AUTOMATIC SPLICES FOR TARGETED ADVERTISEMENTS 审中-公开
    针对广告的自动播放

    公开(公告)号：US20090157499A1

    公开(公告)日：2009-06-18

    申请号：US11957206

    申请日：2007-12-14

    Applicant: Ruston Panabaker ,  John C. Krumm ,  Jeffrey D. Couckuyt ,  Ivan J. Tashev ,  Michael Lewis Seltzer ,  Neil W. Black

    Inventor： Ruston Panabaker ,  John C. Krumm ,  Jeffrey D. Couckuyt ,  Ivan J. Tashev ,  Michael Lewis Seltzer ,  Neil W. Black

    IPC: G06F19/00

    CPC classification number: G06Q30/02 ,  G06Q30/0273

    Abstract: Commonly targeted advertising is used to attract a specific demographic towards a particular product and/or service. An advertisement can be presented over a vehicle radio that relate to a route while a user is driving. Advertisements can be directed to a user, including such information as a user name, a user's destination, contextual information, etc. Advertisements can be masked to appear as thought they are part of a standard program; for instance, a normal radio advertiser's voice can be used.

    Abstract translation: 通常针对性广告用于吸引特定的人口统计特定的产品和/或服务。 广告可以在用户驾驶时通过与路线相关的车载无线电显示。 广告可以被定向到用户，包括诸如用户名，用户的目的地，上下文信息等的信息。广告可以被掩盖以被认为是他们是标准节目的一部分; 例如，可以使用正常的无线电广播者的语音。

Patent Agency Ranking