公开(公告)号：US20140350938A1

公开(公告)日：2014-11-27

申请号：US14454104

申请日：2014-08-07

申请人： AT&T Intellectual Property I, L.P.

发明人： Horst J. Schroeter

IPC分类号： G10L17/20 ,  G10L17/24 ,  G10L17/00

CPC分类号： G10L17/24 ,  G10L17/005 ,  G10L17/04 ,  G10L17/08 ,  G10L17/20

摘要： Disclosed herein are systems, methods, and tangible computer readable-media for detecting synthetic speaker verification. The method comprises receiving a plurality of speech samples of the same word or phrase for verification, comparing each of the plurality of speech samples to each other, denying verification if the plurality of speech samples demonstrate little variance over time or are the same, and verifying the plurality of speech samples if the plurality of speech samples demonstrates sufficient variance over time. One embodiment further adds that each of the plurality of speech samples is collected at different times or in different contexts. In other embodiments, variance is based on a pre-determined threshold or the threshold for variance is adjusted based on a need for authentication certainty. In another embodiment, if the initial comparison is inconclusive, additional speech samples are received.

摘要翻译： 本文公开了用于检测合成说话人验证的系统，方法和有形计算机可读介质。 所述方法包括：接收用于验证的相同单词或短语的多个语音样本，将所述多个语音样本中的每一个相互比较，拒绝验证所述多个语音样本是否随时间表现出很小的变化或相同，并验证 多个语音样本如果多个语音样本显示出随时间的足够的方差。 一个实施例进一步补充说，在不同的时间或在不同的上下文中收集多个语音样本中的每一个。 在其他实施例中，方差基于预定阈值，或者基于对认证确定性的需要来调整方差阈值。 在另一个实施例中，如果初始比较是不确定的，则接收附加语音样本。

公开(公告)号：US20140278412A1

公开(公告)日：2014-09-18

申请号：US13838512

申请日：2013-03-15

申请人： SRI INTERNATIONAL

发明人： NICOLAS SCHEFFER ,  LUCIANA FERRER

IPC分类号： G10L15/06

CPC分类号： G10L25/27 ,  G10L17/20 ,  G10L25/03

摘要： Characterizing an acoustic signal includes extracting a vector from the acoustic signal, where the vector contains information about the nuisance characteristics present in the acoustic signal, and computing a set of likelihoods of the vector for a plurality of classes that model a plurality of nuisance characteristics. Training a system to characterize an acoustic signal includes obtaining training data, the training data comprising a plurality of acoustic signals, where each of the plurality of acoustic signals is associated with one of a plurality of classes that indicates a presence of a specific type of nuisance characteristic, transforming each of the plurality of acoustic signals into a vector that summarizes information about the acoustic characteristics of the signal, to produce a plurality of vectors, and labeling each of the plurality of vectors with one of the plurality of classes.

摘要翻译： 表征声学信号包括从声学信号中提取矢量，其中矢量包含关于存在于声学信号中的扰动特性的信息，以及计算针对多个妨扰特性建模的多个等级的矢量的一组似然性。 训练用于表征声信号的系统包括获得训练数据，所述训练数据包括多个声信号，其中所述多个声信号中的每一个与所述多个类中的一个相关联，所述多个等级中的一个指示存在特定类型的滋扰 将所述多个声信号中的每一个变换为总结关于所述信号的声学特性的信息的矢量，以产生多个向量，并且使用所述多个类之一来标记所述多个向量中的每一个。

公开(公告)号：US20130253920A1

公开(公告)日：2013-09-26

申请号：US13843935

申请日：2013-03-15

申请人： Qiguang LIN

发明人： Qiguang LIN

IPC分类号： G10L17/20

CPC分类号： G10L17/20 ,  G10L15/02 ,  G10L15/20 ,  G10L17/02

摘要： A method of processing a speech signal comprises converting the speech signal to digital signals, converting the digital speech signal into short-time frames, applying a Fast Fourier Transform to each of the short-time frames to obtain an original spectrum, deriving a varied spectrum based on the original spectrum, applying discrete cosine transform to compute original cepstrum coefficients for the original spectrum and varied cepstrum coefficients for the varied spectrum and generating a set of frontend feature vectors for each of the short-time frames.

摘要翻译： 一种处理语音信号的方法包括将语音信号转换为数字信号，将数字语音信号转换为短时帧，将快速傅立叶变换应用于每个短时帧以获得原始频谱，导出不同频谱 基于原始频谱，使用离散余弦变换来计算原始频谱的原始倒频谱系数和用于各种频谱的变化的倒谱系数，并为每个短时间帧生成一组前端特征向量。

公开(公告)号：US08259954B2

公开(公告)日：2012-09-04

申请号：US11870808

申请日：2007-10-11

申请人： Shmuel Shaffer ,  Mukul Jain ,  Labhesh Patel ,  Sanjeev Kumar

发明人： Shmuel Shaffer ,  Mukul Jain ,  Labhesh Patel ,  Sanjeev Kumar

IPC分类号： H04R29/00 ,  H03G3/20 ,  H04B15/00

CPC分类号： G10L21/0364 ,  G10L17/20

摘要： In one embodiment, one or more users may be participating in a conversation. In one example, a first user may be speaking into a speaker end device and a second user may be listening at a listener end device. The second user may be in an environment where noise may be present. Particular embodiments determine characteristics of the noise at the listener end device. Characteristics of a voice signature for a user speaking with the speaker end device are also determined. Comprehension enhancement of voice signals received from speaker end device is then performed based on characteristics of the noise at the listener end device and characteristics of the voice signature. For example, the signature of the voice signals may be altered to lessen the overlap with the noise.

摘要翻译： 在一个实施例中，一个或多个用户可以参与对话。 在一个示例中，第一用户可以说话到扬声器终端设备中，并且第二用户可能在收听终端设备中收听。 第二用户可能处于可能存在噪声的环境中。 特定实施例确定了听众终端设备处的噪声的特征。 与扬声器终端设备通话的用户的语音签名的特征也被确定。 然后根据听者终端设备的噪声特性和语音签名的特征，对从扬声器终端设备接收到的语音信号进行理解增强。 例如，可以改变语音信号的签名以减少与噪声的重叠。

公开(公告)号：US08229739B2

公开(公告)日：2012-07-24

申请号：US12176668

申请日：2008-07-21

申请人： Tadashi Amada

发明人： Tadashi Amada

IPC分类号： G10L19/00

CPC分类号： G10L17/20 ,  G10L15/20 ,  G10L25/27 ,  G10L2021/02165

摘要： A speech processing apparatus includes a plurality of microphones which receive speech produced by a first sound source to obtain first speech signals for a plurality of channels having one-to-one correspondence with the plurality of microphones, a calculation unit configured to calculate a first characteristic amount indicative of an inter-channel correlation of the first speech signals, a storage unit configured to store in advance a second characteristic amount indicative of an inter-channel correlation of second speech signals for the plurality of channels obtained by receiving speech produced by a second sound source by the plurality of microphones, and a collation unit configured to collate the first characteristic amount with the second characteristic amount to determine whether the first sound source matches with the second sound source.

摘要翻译： 语音处理装置包括：多个麦克风，其接收由第一声源产生的语音，以获得与多个麦克风一一对应的多个频道的第一语音信号;计算单元，被配置为计算第一特征 指示第一语音信号的信道间相关的量;存储单元，被配置为预先存储指示通过接收由第二语音信号产生的语音获得的多个频道获得的多个频道的第二语音信号的信道间相关性的第二特征量 多个麦克风的声源，以及对照单元，被配置为将第一特征量与第二特征量进行比较，以确定第一声源是否与第二声源匹配。

公开(公告)号：US20110202340A1

公开(公告)日：2011-08-18

申请号：US13126859

申请日：2009-10-29

申请人： Aladdin M. Ariyaeeinia ,  Surosh G Pillay ,  Mark Pawlewski

发明人： Aladdin M. Ariyaeeinia ,  Surosh G Pillay ,  Mark Pawlewski

IPC分类号： G10L15/20

CPC分类号： G10L17/12 ,  G10L17/20

摘要： A speaker verification method is proposed that first builds a general model of user utterances using a set of general training speech data. The user also trains the system by providing a training utterance, such as a passphrase or other spoken utterance. Then in a test phase, the user provides a test utterance which includes some background noise as well as a test voice sample. The background noise is used to bring the condition of the training data closer to that of the test voice sample by modifying the training data and a reduced set of the general data, before creating adapted training and general models. Match scores are generated based on the comparison between the adapted models and the test voice sample, with a final match score calculated based on the difference between the match scores. This final match score gives a measure of the degree of matching between the test voice sample and the training utterance and is based on the degree of matching between the speech characteristics from extracted feature vectors that make up the respective speech signals, and is not a direct comparison of the raw signals themselves. Thus, the method can be used to verify a speaker without necessarily requiring the speaker to provide an identical test phrase to the phrase provided in the training sample.

摘要翻译： 提出了一种说话人验证方法，其首先使用一组一般训练语音数据构建用户话语的一般模型。 用户还通过提供训练话语来训练系统，例如口令或其他口语说话。 然后在测试阶段，用户提供测试话语，其包括一些背景噪声以及测试语音样本。 背景噪声用于在创建适应的训练和一般模型之前，通过修改训练数据和减少的一般数据集，使训练数据的状况更接近于测试语音样本的状态。 基于适应模型和测试语音样本之间的比较产生匹配分数，根据匹配分数之间的差异计算最终匹配分数。 该最终匹配分数给出测试语音样本和训练话语之间的匹配程度的度量，并且基于来自提取的组成各个语音信号的特征向量的语音特征之间的匹配程度，并且不是直接的 原始信号本身的比较。 因此，该方法可用于验证扬声器，而不一定要求扬声器为训练样本中提供的短语提供相同的测试短语。

公开(公告)号：US20070233483A1

公开(公告)日：2007-10-04

申请号：US11416793

申请日：2006-05-02

申请人： Raja Kuppuswamy ,  Christian Pilz

发明人： Raja Kuppuswamy ,  Christian Pilz

IPC分类号： G10L17/00

CPC分类号： G10L17/20

摘要： Example embodiments provide a speaker authentication technology that compensates for mismatches between enrollment process conditions and test process conditions using correction parameters or correction models, which allow for correcting one of the test voice characterizing parameter set and the enrollment voice characterizing parameter set according to a mismatch between the test process conditions and the enrollment process conditions, thereby obtaining values for the test voice characterizing parameter set and the enrollment voice characterizing parameter set that are based on the same or at least similar process conditions. Alternatively, each of the enrollment and test voice characterizing parameter sets may be normalized to predetermined standard process conditions by using the correction parameters or correction models. This abstract is provided to comply with rules requiring an abstract, and it is submitted with the intention that it will not be used to interpret or limit the scope or meaning of the claims.

摘要翻译： 示例性实施例提供了一种扬声器认证技术，其使用校正参数或校正模型来补偿注册过程条件和测试过程条件之间的不匹配，所述校正参数或校正模型允许根据不匹配来校正测试语音特征参数集和注册语音表征参数集中的一个 测试过程条件和注册过程条件，从而获得基于相同或至少相似过程条件的测试语音特征参数集和注册语音表征参数集的值。 或者，可以通过使用校正参数或校正模型将登记和测试语音特征参数集中的每一个归一化为预定的标准处理条件。 提供本摘要以符合要求摘要的规则，并提交其意图是不会用于解释或限制权利要求书的范围或含义。

公开(公告)号：US20070038460A1

公开(公告)日：2007-02-15

申请号：US11199652

申请日：2005-08-09

申请人： Jari Navratil ,  Ganesh Ramaswamy ,  Ran Zilca

发明人： Jari Navratil ,  Ganesh Ramaswamy ,  Ran Zilca

IPC分类号： G10L11/00

CPC分类号： G10L17/20 ,  G10L17/06

摘要： A system and method for identifying an individual includes collecting biometric information for an individual attempting to gain access to a system. The biometric information for the individual is scored against pre-trained imposter models. If a score is greater than a threshold, the individual as an imposter is identified as an imposter. Other systems and methods are also disclosed.

摘要翻译： 用于识别个体的系统和方法包括收集试图访问系统的个体的生物特征信息。 个人的生物特征信息对预先训练的冒牌者模型进行评分。 如果一个分数大于一个门槛，个人作为冒名顶替者被确定为冒名顶替者。 还公开了其它系统和方法。

公开(公告)号：US06990446B1

公开(公告)日：2006-01-24

申请号：US09685534

申请日：2000-10-10

申请人： Xuedong Huang ,  Michael D. Plumpe

发明人： Xuedong Huang ,  Michael D. Plumpe

IPC分类号： G10L15/08

CPC分类号： G10L17/20 ,  G10L17/04 ,  G10L21/0216

摘要： A method and apparatus for speaker recognition is provided that matches the noise in training data to noise in testing data using spectral addition. Under spectral addition, the mean and variance for a plurality of frequency components are adjusted in the training data and the test data so that each mean and variance is matched in a resulting matched training signal and matched test signal. The adjustments made to the training data and test data add to the mean and variance of the training data and test data instead of subtracting from the mean and variance.

摘要翻译： 提供一种用于说话者识别的方法和装置，其使用频谱添加将训练数据中的噪声与测试数据中的噪声相匹配。 在光谱加法下，在训练数据和测试数据中调整多个频率分量的平均值和方差，使得​​所得到的匹配训练信号和匹配测试信号中的每个平均值和方差相匹配。 对训练数据和测试数据进行的调整增加了训练数据和测试数据的均值和方差，而不是从均值和方差中减去。

公开(公告)号：US20040236573A1

公开(公告)日：2004-11-25

申请号：US10481523

申请日：2004-06-16

发明人： Andrew Thomas Sapeluk

IPC分类号： G10L019/14

CPC分类号： G10L17/02 ,  G10L17/12 ,  G10L17/20

摘要： Speaker recognition (identification and/or verification) methods and systems, in which speech models for enrolled speakers consist of sets of feature vectors representing the smoothed frequency spectrum of each of a plurality of frames and a clustering algorithm is applied to the feature vectors of the frames to obtain a reduced data set representing the original speech sample, and wherein the adjacent frames are overlapped by at least 80%. Speech models of this type model the static components of the speech sample and exhibit temporal independence. An identifier strategy is employed in which modelling and classification processes are selected to give a false rejection rate substantially equal to zero. Each enrolled speaker is associated with a cohort of a predetermined number of other enrolled speakers and a test sample is always matched with either the claimed identity or one of its associated cohort. This makes the overall error rate of the system dependent only on the false acceptance rate, which is determined by the cohort size. The false error rate is further reduced by use of multiple parallel modelling and/or classification processes. Speech models are normalised prior to classification using a normalisation model derived from either the test speech sample or one of the enrolled speaker samples (most preferably from the claimed identity enrolment sample).

摘要翻译： 扬声器识别（识别和/或验证）方法和系统，其中登记的扬声器的语音模型由表示多个帧中的每一个的平滑频谱的特征向量集合和聚类算法应用于 帧以获得表示原始语音样本的缩减数据集，并且其中相邻帧重叠至少80％。 这种类型的语音模型模拟语音样本的静态组件并呈现时间独立性。 采用标识符策略，其中选择建模和分类处理以给出基本等于零的错误拒绝率。 每个登记的说话者与预定数量的其他注册的发言人的队列相关联，并且测试样本总是与所要求保护的身份或其相关联的队列中的一个匹配。 这使得系统的总体错误率仅取决于由队列大小确定的错误接受率。 通过使用多个并行建模和/或分类过程进一步降低了错误错误率。 语音模型在使用从测试语音样本或所登记的说话者样本（最优选来自所要求的身份登记样本）导出的归一化模型之前进行归一化。