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. 发明专利

JP2013167698A Apparatus and method for estimating spectral shape feature quantity of signal for every sound source, and apparatus, method and program for estimating spectral feature quantity of target signal 有权
标题翻译：用于估计每个声源的光谱特征信号量的装置和方法，以及用于估计目标信号的光谱特征量的装置，方法和程序
公开(公告)号：JP2013167698A
公开(公告)日：2013-08-29
申请号：JP2012029791
申请日：2012-02-14
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： NAKATANI TOMOHIRO , YOSHIOKA TAKUYA , ARAKI AKIKO , DELCROIX MARC , FUJIMOTO MASAKIYO
IPC分类号： G10L25/27 , G10L15/02 , G10L21/0308 , G10L25/24
摘要： PROBLEM TO BE SOLVED: To provide a technique for estimating a spectral shape feature quantity of a signal for every sound source, thereby efficiently performing a spectral estimation of a target sound, even using a model in which spectral values have correlation between frequencies.SOLUTION: A prior probability density function (a spectral shape model) of a spectral shape feature quantity corresponding to each sound source, and a conditioned probability density function (a spectral observation model) of the spectral feature quantity when the spectral shape feature quantity is given are used. An optimization function is represented by a product of the conditioned probability density function of the spectral shape feature quantity which has, in a latent variable, an occupancy sound source number representing a sound source of an acoustic signal having maximal energy in each time frequency point and to which the spectral shape feature quantities of all sound sources are given, and the prior probability density function of the spectral shape feature quantity determined for every sound source. The optimization function is maximized by using the spectral shape model and the spectral observation model, to estimate the spectral shape feature quantity for every sound and a sound source occupancy.
摘要翻译：要解决的问题：提供一种用于估计每个声源的信号的频谱形状特征量的技术，从而即使使用其中频谱值具有频率之间的相关性的模型，也可以有效地执行目标声音的频谱估计。解决方案：使用对应于每个声源的光谱形状特征量的先验概率密度函数（光谱形状模型）和当给出光谱形状特征量时的光谱特征量的调节概率密度函数（光谱观察模型）。优化函数由频谱形状特征量的条件概率密度函数的乘积表示，在潜变量中，具有表示每个时间频点具有最大能量的声信号的声源的占用声源数，以及给出了所有声源的光谱形状特征量的特征量，并且为每个声源确定了光谱形状特征量的先验概率密度函数。通过使用光谱形状模型和光谱观测模型，优化功能最大化，以估计每个声音和声源占用的光谱形状特征量。

2. 发明专利

JP2013044909A Background sound suppressor, background sound suppression method and program 有权
标题翻译：声音抑制器，背景声音抑制方法和程序
公开(公告)号：JP2013044909A
公开(公告)日：2013-03-04
申请号：JP2011182278
申请日：2011-08-24
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： NAKATANI TOMOHIRO , ARAKI AKIKO , YOSHIOKA TAKUYA , FUJIMOTO MASAKIYO , DELCROIX MARC
IPC分类号： G10L21/0232 , G10L15/20 , G10L21/0208
摘要： PROBLEM TO BE SOLVED: To provide a more efficient and highly accurate background sound suppressor capable of reducing a calculation cost and utilizing a probability density function of more complex form.SOLUTION: In a background sound suppressor 40 of the present invention, a feature quantity extraction unit 100 extracts a high resolution sound source position feature quantity and a high resolution spectral feature quantity from an observation signal, a sound source position occupation degree estimation unit 210 obtains a high resolution sound source position occupation degree, a high resolution occupation degree estimation unit 520 estimates a high resolution occupation degree and a spectral parameter, and a target speech estimation unit 600 estimates a target speech.
摘要翻译：要解决的问题：提供一种能够降低计算成本并利用更复杂形式的概率密度函数的更有效和高度准确的背景声音抑制器。解决方案：在本发明的背景抑制器40中，特征量提取单元100从观测信号，声源位置占用度估计中提取高分辨率声源位置特征量和高分辨率光谱特征量单元210获得高分辨率声源位置占有度，高分辨率占用度估计单元520估计高分辨率占用程度和频谱参数，并且目标语音估计单元600估计目标语音。版权所有（C）2013，JPO＆INPIT

3. 发明专利

JP2013044908A Background sound suppressor, background sound suppression method and program 有权
标题翻译：声音抑制器，背景声音抑制方法和程序
公开(公告)号：JP2013044908A
公开(公告)日：2013-03-04
申请号：JP2011182277
申请日：2011-08-24
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： NAKATANI TOMOHIRO , ARAKI AKIKO , YOSHIOKA TAKUYA , FUJIMOTO MASAKIYO , DELCROIX MARC
IPC分类号： G10L21/0232
摘要： PROBLEM TO BE SOLVED: To provide a more efficient and highly accurate background sound suppressor capable of reducing a calculation cost and utilizing a probability density function with a more complex form.SOLUTION: In a background sound suppressor 20 of the present invention, a feature quantity extraction unit 100 extracts a high resolution sound source position feature quantity and a high resolution spectral feature quantity from an observation signal, a sound source position occupation degree estimation unit 200 obtains a high resolution sound source position occupation degree, a frequency resolution reduction unit 300 reduces frequency resolution of the high resolution spectral feature quantity and the high resolution sound source position occupation degree, a low resolution occupation degree estimation unit 400 estimates a spectral parameter, a high resolution occupation degree re-estimation unit 510 estimates a high resolution occupation degree, and a target speech estimation unit 600 estimates a target speech.
摘要翻译：要解决的问题：提供能够降低计算成本并利用更复杂形式的概率密度函数的更有效和高度准确的背景声音抑制器。解决方案：在本发明的背景声音抑制器20中，特征量提取单元100从观测信号，声源位置占用度估计中提取高分辨率声源位置特征量和高分辨率光谱特征量单元200获得高分辨率声源位置占有度，频率分辨率降低单元300降低高分辨率频谱特征量和高分辨率声源位置占有度的频率分辨率，低分辨率占用度估计单元400估计频谱参数高分辨率占有度重新估计单元510估计高分辨率占用程度，并且目标语音估计单元600估计目标语音。版权所有（C）2013，JPO＆INPIT

4. 发明专利

JP2014153680A Acoustic model correction parameter estimation device, feature quantity correction parameter estimation device, and methods and programs therefor 有权
标题翻译：声学模型校正参数估计装置，特征量子校正参数估计装置及其方法和程序
公开(公告)号：JP2014153680A
公开(公告)日：2014-08-25
申请号：JP2013025865
申请日：2013-02-13
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： DELCROIX MARC , OGAWA ATSUNORI , HAM SONJUN , NAKATANI TOMOHIRO , NAKAMURA ATSUSHI
IPC分类号： G10L15/065 , G10L15/06 , G10L15/14
摘要： PROBLEM TO BE SOLVED: To provide an acoustic model correction parameter estimation technique and a feature quantity correction parameter estimation technique that prevent acoustic model adaptation from decreasing in precision even in unsupervised adaptation in which there are many errors in correct answer symbol by introducing a mechanism for reducing an adverse effect of an error in correct answer symbol.SOLUTION: An acoustic model correction parameter learning device 100 includes: a correction parameter differential value calculation part 140 which finds a differential value, obtained when an object function for discrimination learning standard is differentiated with an average correction parameter, based upon language probability of an opposing candidate symbol series obtained with a language model, an acoustic score obtained with an acoustic model based upon a feature quantity of voice data for learning and the opposing candidate symbol series, and dissimilarity; and a correction parameter update part 150 which updates the average correction parameter by changing the average correction parameter according to the differential value.
摘要翻译：要解决的问题：提供一种声学模型校正参数估计技术和特征量校正参数估计技术，其特征量校正参数估计技术，即使在无监督自适应中也能防止声学模型适应降低，其中在正确答案符号中存在许多误差，减少正确答案符号中误差的不利影响。解决方案：声学模型校正参数学习装置100包括：校正参数差分值计算部分140，其找到当鉴别学习标准的对象功能与基于用语言模型获得的相对候选符号序列的语言概率的平均校正参数，基于用于学习的语音数据的特征量和相对的候选符号序列的声学模型获得的声学得分和相异性; 以及校正参数更新部150，其通过根据差分值改变平均校正参数来更新平均校正参数。

5. 发明专利

JP2013174769A Dispersion correction parameter estimation device, voice recognition system, dispersion correction parameter estimation method, voice recognition method and program 有权
标题翻译：分散校正参数估计装置，语音识别系统，分散校正参数估计方法，语音识别方法和程序
公开(公告)号：JP2013174769A
公开(公告)日：2013-09-05
申请号：JP2012039819
申请日：2012-02-27
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： DELCROIX MARC , OGAWA ATSUNORI , NAKATANI TOMOHIRO , NAKAMURA ATSUSHI
IPC分类号： G10L15/14 , G10L15/06 , G10L15/065
摘要： PROBLEM TO BE SOLVED: To provide a dispersion correction parameter estimation device that stably and accurately estimates a dispersion correction parameter through discriminative learning.SOLUTION: A dispersion correction parameter estimation device executes the steps of: generating a dispersion correction parameter; using the dispersion correction parameter to correct a dispersion parameter of a Gauss distribution included in a mixed Gauss distribution model; obtaining a degree of difference from a correct answer symbol series with a predetermined particle size, with respect to each opposing candidate symbol series to be obtained through voice recognition of a feature quantity of voice data for learning, on the basis of an acoustic model including the corrected dispersion parameter; obtaining a differential value when using the Gauss distribution dispersion correction parameter to differentiate an objective function of a discriminative learning criterion, on the basis of a language probability of the opposing candidate symbol series, an acoustic score to be obtained by an acoustic model on the basis of the feature quantity of voice data for learning and the opposing candidate symbol series and the degree of difference; and updating the Gauss distribution dispersion correction parameter by changing the Gauss distribution correction parameter according to the differential value.
摘要翻译：要解决的问题：提供一种通过识别学习稳定且准确地估计色散校正参数的色散校正参数估计装置。解决方案：色散校正参数估计装置执行以下步骤：产生色散校正参数; 使用色散校正参数来校正包含在混合高斯分布模型中的高斯分布的色散参数; 基于包括该信息的声学模型，获得与通过用于学习的语音数据的特征量的语音识别获得的每个相对的候选符号序列获得具有预定粒度的正确答案符号序列的差异度校正色散参数; 当使用高斯分布色散校正参数来区分鉴别学习标准的目标函数时，基于相对候选符号序列的语言概率，基于声学模型获得的声学得分，获得差分值用于学习的语音数据和相对候选符号序列的特征量和差异度; 以及通过根据所述差分值改变高斯分布校正参数来更新高斯分布色散校正参数。

6. 发明专利

JP2013174768A Feature quantity correction parameter estimation device, voice recognition system, feature quantity correction parameter estimation method, voice recognition method and program 有权
标题翻译：特征量校正参数估计装置，语音识别系统，特征量校正参数估计方法，语音识别方法和程序
公开(公告)号：JP2013174768A
公开(公告)日：2013-09-05
申请号：JP2012039818
申请日：2012-02-27
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： DELCROIX MARC , OGAWA ATSUNORI , NAKATANI TOMOHIRO , NAKAMURA ATSUSHI
IPC分类号： G10L15/20
摘要： PROBLEM TO BE SOLVED: To provide a feature quantity correction parameter estimation device and method that stably estimate a feature quantity correction parameter through discriminative learning, in consideration of a degree of inconsistency between a feature of voice data for learning and that of voice data for recognition.SOLUTION: A feature quantity correction parameter estimation device includes: a correction parameter differential value calculation unit that obtains a differential value when using a feature quantity correction parameter to differentiate an objective function of dMMI discriminative learning criterion, on the basis of a language probability of an opposing candidate symbol series to be obtained by a language model, an acoustic score to be obtained by an acoustic model on the basis of the corrected feature quantity and the opposing candidate symbol series, a first margin parameter to be adjusted according to a degree of inconsistency between a feature of voice data for learning and that of voice data for recognition, a second margin parameter and a degree of difference; and a correction parameter update unit that updates the feature quantity correction parameter by changing the feature quantity correction parameter according to the differential value.
摘要翻译：要解决的问题：考虑到用于学习的语音数据的特征与用于识别的语音数据的特征之间的不一致性，提供了通过鉴别学习来稳定地估计特征量校正参数的特征量校正参数估计装置和方法解决方案：特征量校正参数估计装置包括：校正参数差分值计算单元，其基于特征量校正参数来获取差分值，以根据dMMI鉴别学习标准的目标函数来区分目标函数，通过语言模型获得的相对候选符号序列，通过基于校正特征量和相对候选符号序列的声学模型获得的声学得分，根据程度语音数据特征之间的不一致 r学习和语音数据识别，第二个边距参数和程度的差异; 以及校正参数更新单元，其通过根据差分值改变特征量校正参数来更新特征量校正参数。

7. 发明专利

JP2013037177A Speech enhancement device, and method and program thereof 有权
标题翻译：语音增强设备及其方法和程序
公开(公告)号：JP2013037177A
公开(公告)日：2013-02-21
申请号：JP2011172939
申请日：2011-08-08
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： MERETSU SODEN , KINOSHITA KEISUKE , NAKATANI TOMOHIRO , DELCROIX MARC
IPC分类号： G10L21/0216 , G10L21/0208
摘要： PROBLEM TO BE SOLVED: To provide a speech enhancement device for which space information is available.SOLUTION: A feature quantity estimation part regards a multichannel sound signal as observation signal and extracts feature quantities of sound signals of respective channels, and an acoustic strain parameter estimation part uses a feature quantity of a multichannel observation signal and a clean speech Gaussian mixed model as input and performs maximum likelihood estimation of an acoustic strain parameter relating to an additive noise component included in the observation signal and a channel response strain. Then, a clean voice feature quantity estimation part estimates a feature quantity of a clean speech signal by using the feature quantity of a multichannel observation signal, the acoustic strain parameter estimated in the acoustic strain parameter estimation part, and the clean speech Gaussian mixed model as input and by performing least square error estimation of the clean speech signal included in the observation signal.
摘要翻译：要解决的问题：提供可用空间信息的语音增强装置。解决方案：特征量估计部分将多声道声音信号作为观测信号，并提取各声道的声音信号的特征量，并且声应变参数估计部分使用多声道观测信号和清晰语音高斯的特征量混合模型作为输入，并且执行与包括在观察信号中的加性噪声分量有关的声应变参数和信道响应应变的最大似然估计。然后，干净语音特征量估计部通过使用多声道观测信号的特征量，在声学应变参数估计部中估计的声应变参数和清晰语音高斯混合模型来估计干净语音信号的特征量，作为并且通过对包括在观察信号中的清洁语音信号执行最小二乘误差估计。版权所有（C）2013，JPO＆INPIT

8. 发明专利

JP2013037174A Noise/reverberation removal device, method thereof, and program 有权
标题翻译：噪声/反复删除设备，其方法和程序
公开(公告)号：JP2013037174A
公开(公告)日：2013-02-21
申请号：JP2011172919
申请日：2011-08-08
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： KINOSHITA KEISUKE , NAKATANI TOMOHIRO , MERETSU SODEN , DELCROIX MARC
IPC分类号： G10L21/0208 , G10L15/20 , G10L21/0232
摘要： PROBLEM TO BE SOLVED: To provide a noise/reverberation removal device which uses only exemplar models of clean voice to perform voice emphasis.SOLUTION: An emphasizing processing result reliability calculation unit outputs a value indicative of uncertainty of a primary voice emphasis signal in accordance with a feature quantity of an input signal and the primacy voice emphasis signal. A matching unit receives the primary voice emphasis signal, the value indicative of uncertainty of the primary voice emphasis signal, and exemplar models of learning data and outputs a learning data segment which gives a clean voice sequence closest to clean voice included in the input signal with respect to each time frame. A voice emphasis filtering unit receives the input signal and the learning data segment, reads out amplitude spectrum data pairing the learning data segment from an exemplar model storage unit to generate a Wiener filter, multiplies a power spectrum of the input signal by the Wiener filter to perform filtering, and outputs a voice emphasis signal.
摘要翻译：要解决的问题：提供仅使用干净声音的示范模型来执行语音强调的噪声/混响去除装置。解决方案：强调处理结果可靠性计算单元根据输入信号的特征量和首要语音强调信号，输出表示主要语音强调信号的不确定性的值。匹配单元接收主要语音重点信号，指示主要语音重点信号的不确定性的值以及学习数据的示例模型，并输出学习数据段，该学习数据段给包括在输入信号中的干净语音提供干净的语音序列，尊重每个时间框架。语音强调滤波单元接收输入信号和学习数据段，从样本模型存储单元中读出配对学习数据段的振幅谱数据，生成维纳滤波器，将维纳滤波器的输入信号的功率谱乘以执行滤波，并输出语音强调信号。版权所有（C）2013，JPO＆INPIT

9. 发明专利

JP2009145499A Voice parameter learning apparatus and method therefor, voice recognition apparatus and voice recognition method using them, and their program and recording medium 有权
标题翻译：语音参数学习装置及其方法，语音识别装置和使用它们的语音识别方法及其程序和记录介质
公开(公告)号：JP2009145499A
公开(公告)日：2009-07-02
申请号：JP2007321201
申请日：2007-12-12
申请人： Nippon Telegr & Teleph Corp , 日本電信電話株式会社
发明人： DELCROIX MARC , WATABE SHINJI , NAKATANI TOMOHIRO
IPC分类号： G10L15/06 , G10L15/14 , G10L15/20
摘要： PROBLEM TO BE SOLVED: To provide a voice-parameter learning apparatus that does not depend on specific voice emphasis method.
SOLUTION: This voice-parameter learning apparatus includes a voice preprocessing section for adaptation, an acoustic model storage section, an adaptation parameter creating section, a voice preprocessing section for recognition, and a dispersion dynamic correcting section. The adaptation parameter creating section creates a dynamic dispersion adaptive parameter depending on a frame as a parameter for dispersion correction, and a static dispersion adaptive parameter independent of the frame. The voice preprocessing section for recognition creates a voice feature amount for each frame of an observation voice signal, and uncertainty showing variation in the voice feature amount. The dispersion dynamic correcting section receives the uncertainty of the voice feature amount, the adaptive parameter, and the acoustic model, and outputs the dispersion of the Gaussian distribution corrected with the adaptive parameter for each frame.
COPYRIGHT: (C)2009,JPO&INPIT
摘要翻译：要解决的问题：提供不依赖于特定语音强调方法的语音参数学习装置。解决方案：该语音参数学习装置包括用于适应的语音预处理部分，声学模型存储部分，适应参数创建部分，用于识别的语音预处理部分和分散动态校正部分。适应参数创建部分根据作为色散校正的参数的帧创建动态色散自适应参数，以及独立于帧的静态色散自适应参数。用于识别的语音预处理部分创建观察语音信号的每个帧的语音特征量，以及示出语音特征量的变化的不确定性。分散动态校正部分接收语音特征量的不确定性，自适应参数和声学模型，并且输出用于每个帧用自适应参数校正的高斯分布的色散。版权所有（C）2009，JPO＆INPIT

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