公开(公告)号：KR1020110024969A

公开(公告)日：2011-03-09

申请号：KR1020090083167

申请日：2009-09-03

Applicant: 한국전자통신연구원

Inventor： 이성주 ,  이윤근 ,  강병옥 ,  강점자 ,  박기영 ,  박전규 ,  왕지현 ,  전형배 ,  정의석 ,  정호영 ,  정훈 ,  김종진 ,  박상규

IPC: G10L21/0208 ,  G10L15/20

CPC classification number: G10L21/0208 ,  G10L19/038 ,  G10L19/26

Abstract: PURPOSE: An apparatus for filtering a noise using a statistical model in a voice signal is provided to improve the wiener filter performance by restoring voice signal using a joint density GMM(Gaussian Mixture Model). CONSTITUTION: An apparatus for filtering a noise using a statistical model comprises a clean signal spectrum vector estimating unit(214), a post SNR estimating unit(216), a transfer function estimating unit(218), and a noise filtering unit(220). The clean signal spectrum vector estimating unit estimates spectrum vector of a clean signal using a PSD(Power Spectrum Density), a PSD estimation information of the estimated input signal, and a preset statistical model. The noise filtering unit performs noise filtering using the transfer function and fast fourier transformed frequency axis complex signal.

Abstract translation: 目的：提供一种使用语音信号中的统计模型对噪声进行滤波的装置，以通过使用关节密度GMM（高斯混合模型）恢复语音信号来提高维纳滤波器性能。 构成：使用统计模型对噪声进行滤波的装置包括干净信号频谱矢量估计单元（214），后SNR估计单元（216），传递函数估计单元（218）和噪声滤波单元（220） 。 清洁信号频谱矢量估计单元使用PSD（功率谱密度）估计干净信号的频谱矢量，估计输入信号的PSD估计信息和预设的统计模型。 噪声滤波单元使用传递函数和快速傅立叶变换频率轴复信号进行噪声滤波。

公开(公告)号：KR1020100073172A

公开(公告)日：2010-07-01

申请号：KR1020080131766

申请日：2008-12-22

Applicant: 한국전자통신연구원

Inventor： 왕지현 ,  정의석 ,  강병옥 ,  박전규 ,  강점자 ,  김종진 ,  박기영 ,  이성주 ,  전형배 ,  정호영 ,  정훈 ,  이윤근

IPC: H04N21/232 ,  G10L15/00

CPC classification number: H04N21/232 ,  G06F17/30002 ,  G10L15/02 ,  H04N21/43

Abstract: PURPOSE: A method and a server servicing IPTV broadcast and an IPTV set top apparatus are provided to search the large amount of IPTV broadcasting data by searching broadcasting data in real time through voice speech. CONSTITUTION: A voice recognition list transceiver(103) transmits and receives voice recognition list request information of an update scheduling unit(101). A voice recognition unit(105) recognizes a voice signal in an IPTV set-top box(100), and a voice search request transmitter(106) transmits a recognized string to a voice search request receiver. A voice search result receiver(107) provides the broadcasting data of a broadcasting data transmitter to a display unit(108).

Abstract translation: 目的：提供一种服务于IPTV广播的方法和服务器，以及IPTV机顶装置，通过语音语音实时搜索广播数据来搜索大量的IPTV广播数据。 构成：语音识别列表收发器（103）发送和接收更新调度单元（101）的语音识别列表请求信息。 语音识别单元（105）识别IPTV机顶盒（100）中的语音信号，语音搜索请求发送器（106）将识别的串发送到语音搜索请求接收器。 语音搜索结果接收器（107）将广播数据发送器的广播数据提供给显示单元（108）。

公开(公告)号：KR1020100073167A

公开(公告)日：2010-07-01

申请号：KR1020080131761

申请日：2008-12-22

Applicant: 한국전자통신연구원

Inventor： 박기영 ,  정호영 ,  이윤근 ,  박전규 ,  강점자 ,  정훈 ,  이성주 ,  강병옥 ,  왕지현 ,  정의석 ,  전형배 ,  김종진

IPC: G10L15/20 ,  G10L15/10 ,  G10L21/0272 ,  G10L19/00

CPC classification number: H04R3/005 ,  H04R27/00 ,  H04R2430/03

Abstract: PURPOSE: A method for separating a source signals and an apparatus thereof are provided to improve the recording, transmission and recognition performances by separating only desirable sound source signal in plural sound source environments. CONSTITUTION: Fourier transformer(10) transforms a mixed input signal(S1) into each channel frequency domain through Fourier transformation. A frequency bandwidth divider(20) constitutes a frequency cluster from the each frequency domain. A frequency domain signal divider(30) applies a blind source separation for each cluster frequency domain. A reverse Fourier transformer(40) integrates the spectrums of divided signals through reverse Fourier transformation.

Abstract translation: 目的：提供一种用于分离源信号的方法及其装置，以通过在多个声源环境中分离所需的声源信号来提高记录，传输和识别性能。 构成：傅里叶变换器（10）通过傅里叶变换将混合输入信号（S1）转换成每个通道频域。 频率带宽分配器（20）构成来自每个频域的频率簇。 频域信号分频器（30）为每个群集频域应用盲源分离。 反傅里叶变换器（40）通过反傅里叶变换对分频信号的频谱进行积分。

公开(公告)号：KR1020100072842A

公开(公告)日：2010-07-01

申请号：KR1020080131369

申请日：2008-12-22

Applicant: 한국전자통신연구원

Inventor： 이성주 ,  정호영 ,  박전규 ,  정훈 ,  이윤근 ,  강병옥 ,  전형배 ,  김종진 ,  박기영 ,  정의석 ,  왕지현 ,  강점자

IPC: G10L21/02 ,  G10L15/20 ,  G10L15/14 ,  G10L15/28

CPC classification number: G10L21/0208 ,  G10L15/20 ,  G10L25/48

Abstract: PURPOSE: A speech improving apparatus and a speech recognition system and method are provided to improve the voice recognition performance of a voice recognition system in a movable body of small resources by performing signal decoding through a sound model database. CONSTITUTION: A speed level divider(100) measures a moving speed level of a movable body through an inputted noise signal inputted in an initial stage of voice recognition. When the speed level of the movable body is lower than a predetermined value, a first sound quality improvement unit(112) improves the sound quality of a voice signal inputted by a Wiener filter. If the speed level of the movable body exceeds a predetermined value, a second sound quality improvement unit(114) improves the sound quality of a voice signal inputted by a GMM(Gaussian Mixture Model).

Abstract translation: 目的：提供语音改善装置和语音识别系统和方法，通过声音模型数据库执行信号解码来提高小资源移动体中语音识别系统的语音识别性能。 构成：速度分级器（100）通过在语音识别的初始阶段输入的输入噪声信号测量可移动体的移动速度水平。 当可移动体的速度水平低于预定值时，第一音质改善单元（112）提高了由维纳滤波器输入的语音信号的声音质量。 如果可移动体的速度水平超过预定值，则第二音质改善单元（114）提高了由GMM（高斯混合模型）输入的语音信号的声音质量。

公开(公告)号：KR1020100069117A

公开(公告)日：2010-06-24

申请号：KR1020080127707

申请日：2008-12-16

Applicant: 한국전자통신연구원

Inventor： 전형배 ,  정호영 ,  박전규 ,  정훈 ,  이윤근 ,  강점자 ,  정의석 ,  강병옥 ,  김종진 ,  왕지현 ,  이성주 ,  박기영

IPC: G10L25/78 ,  G10L25/24 ,  G10L15/14 ,  G10L15/20

Abstract: PURPOSE: A CMS(Cepstrum Mean Subtraction) method and a device thereof are provided to accurately normalize a channel property by estimating an average CMS value of the real voice section based on the CMS average value of a mute section. CONSTITUTION: A property extractor(200) extracts the properties of a mute section before a start point, a sound section, and a mute section after a finish point. A firing unit CMS value calculator(600) calculates an actual firing unit cepstrum average about the entire sound section. A cepstrum average estimator(300) estimates the cepstrum average of the entire section based on the properties of the mute section. A property vector CMS applier(400) performs channel-normalization of the estimated average. A decoder decodes the channel-normalized MFCC property vector.

Abstract translation: 目的：提供CMS（倒谱平均减法）方法及其装置，以通过基于静音部分的CMS平均值估计真实语音部分的平均CMS值来准确地规范信道特性。 规定：属性提取器（200）在完成点之后提取起始点，声音部分和静音部分之前的静音部分的属性。 点火单元CMS值计算器（600）计算关于整个声音部分的实际发射单位倒谱平均值。 倒谱平均估计器（300）基于静音部分的属性来估计整个部分的倒谱平均值。 属性向量CMS应用程序（400）执行估计平均值的信道归一化。 解码器解码信道归一化的MFCC属性向量。

公开(公告)号：KR1020100066917A

公开(公告)日：2010-06-18

申请号：KR1020080125434

申请日：2008-12-10

Applicant: 한국전자통신연구원

Inventor： 정의석 ,  왕지현 ,  강병옥 ,  박전규 ,  강점자 ,  김종진 ,  박기영 ,  이성주 ,  전형배 ,  정호영 ,  정훈 ,  이윤근

IPC: G01C21/34 ,  G01C21/36 ,  G01C21/32 ,  G01C21/00

CPC classification number: G01C21/3608 ,  G01C21/3611 ,  G01C21/3629 ,  G06F17/3074 ,  G10L15/18 ,  G10L15/22

Abstract: PURPOSE: A voice recognition method of a vehicle navigation terminal is provided to generate voice emitting isoform through a simple pattern construction using a resolute/tagged result by presenting a meaning classification system for POI name domain. CONSTITUTION: A voice recognition method of a vehicle navigation terminal is as follows. The points of interest(POI) list and POI learning data are recognized from the voice information of a voice emitting isoform input to the vehicle navigation terminal (S200). A resource is built on the POI list and the POI learning data recognized(S202). The resolution and tagging on the built resource are performed with the POI list(S204). The result resolved and tagged is created as POI database(S206). Simplex/analyzed database is built based on the POI list and the POI learning data. N-gram vocabulary is extracted from the POI learning data.

Abstract translation: 目的：提供一种车载导航终端的语音识别方法，通过呈现POI名称域的意义分类系统，通过简单的模式构造，通过坚决/标记的结果生成语音发射同种型。 构成：车辆导航终端的声音识别方法如下。 通过输入到车辆导航终端的发音同步体的语音信息来识别兴趣点（POI）列表和POI学习数据（S200）。 资源建立在POI列表和POI学习数据识别（S202）上。 使用POI列表执行内置资源的分辨率和标记（S204）。 解决和标记的结果被创建为POI数据库（S206）。 基于POI列表和POI学习数据构建S​​implex /分析数据库。 从POI学习数据中提取N-gram词汇表。

公开(公告)号：KR1020090111739A

公开(公告)日：2009-10-27

申请号：KR1020080075653

申请日：2008-08-01

Applicant: 한국전자통신연구원

Inventor： 강병옥 ,  정호영 ,  이성주 ,  이윤근 ,  박전규 ,  강점자 ,  정훈 ,  정의석 ,  왕지현 ,  전형배

IPC: G10L15/20

Abstract: PURPOSE: A noise cancelling apparatus is provided to estimate a clean voice more accurately in an environment in which a dynamic noise and various noises are mixed. CONSTITUTION: A noise cancelling apparatus comprises a noise estimation module(200) which calculates the estimation value of a noise signal in the current frame of a voice signal, a Wiener filter module(202) which receives the voice signal and calculates an intermediate result by applying the intermediate Wiener filter, a database(206) in which Gaussian mixed-model data is stored, and an MMSE estimation module(204) which calculates the estimation value of a clean voice by using the Gaussian mixed-model data and intermediate result.

Abstract translation: 目的：提供一种噪声消除装置，用于在动态噪声和各种噪声混合的环境中更精确地估计干净的声音。 噪声消除装置包括噪声估计模块（200），噪声估计模块（200），其计算语音信号的当前帧中的噪声信号的估计值;维纳滤波器模块（202），其接收语音信号并通过以下步骤计算中间结果 应用中间维纳滤波器，存储高斯混合模型数据的数据库（206）和通过使用高斯混合模型数据和中间结果来计算干净声音的估计值的MMSE估计模块（204）。

公开(公告)号：KR1020080021486A

公开(公告)日：2008-03-07

申请号：KR1020070066452

申请日：2007-07-03

Applicant: 한국전자통신연구원

Inventor： 이창기 ,  이충희 ,  오효정 ,  왕지현 ,  김현진 ,  황이규 ,  장명길 ,  이영직

IPC: G10L15/02 ,  G10L15/04 ,  G10L15/06 ,  G10L15/22

Abstract: An apparatus and a method for generating a response sentence are provided to perform exact meaning analysis of a voice-recognized sentence by performing second point of sentence/substitutes extraction and second meaning analysis with respect to the voice-recognized sentence. A response sentence generating method comprises the following steps of: performing morpheme analysis of a voice-recognized sentence(200,210); extracting a first point of sentence from the sentence(220); performing first meaning analysis of the sentence based on the extracted first point of sentence(230); extracting a second point of sentence including the first point of sentence from the sentence based on the first meaning analysis result in order to further extract point of sentence which are not extracted in the above second step(240); generating a meaning analysis result of the voice-recognized sentence by performing second meaning analysis of the sentence based on the extracted second point of sentence(250); and generating a response sentence to the voice-recognized sentence based on the generated meaning analysis result(260).

Abstract translation: 提供了一种用于产生响应语句的装置和方法，以通过执行语音识别句子的句子/替换提取和第二含义分析的第二点来执行语音识别句子的精确含义分析。 响应语句生成方法包括以下步骤：对语音识别句子进行语素分析（200,210）; 从句子中提取第一句话（220）; 基于提取的第一句话（230），对句子进行第一意义分析; 为了进一步提取在上述第二步骤（240）中未被提取的句子，提取基于第一意义分析结果从句子中提取包括第一句子的句子的第二句子; 通过基于提取的第二句子点（250）对句子进行第二含义分析，从而产生语音识别句子的意义分析结果; 以及基于所生成的含义分析结果（260）生成对所述语音识别句子的响应句子。

公开(公告)号：KR1020040050461A

公开(公告)日：2004-06-16

申请号：KR1020020078305

申请日：2002-12-10

Applicant: 한국전자통신연구원

Inventor： 정의석 ,  왕지현 ,  윤보현

IPC: G06F17/27

Abstract: PURPOSE: A device and a method for recognizing an object name on a Korean text are provided to recognize the object name on the Korean text through the reinforced learning using a co-training based on the HMM(Hidden Markov Model). CONSTITUTION: A morpheme parser(10) separates the input text into a list of sentences, forms a morpheme list separating each sentence into the morpheme unit tagging a state label, generates/stores an HMM data structure in a memory. A statistics information extractor(20) extracts the HMM object name statistics information from an object name tagged text set. A co-training learning device(40) extends the statistics information through the current learning data by extracting the HMM object name statistics information from an unlabeled text set as advancing the co-training learning based on the HMM statistics information extracted from the object name tagged text set. An object name recognizer(50) recognizes the object name by deciding an optimal HMM object name statistics information path of the morphemes forming the sentences of the input text through a viterbi algorithm.

Abstract translation: 目的：提供用于识别韩文文本中的对象名称的设备和方法，以通过使用基于HMM（隐马尔可夫模型）的共同训练的加强学习来识别韩文文本上的对象名称。 构成：词素解析器（10）将输入的文本分成句子列表，形成一个语素列表，将每个句子分离成标签状态标签的语素单元，将HMM数据结构生成/存储在存储器中。 统计信息提取器（20）从对象名称标记的文本集中提取HMM对象名称统计信息。 共同训练学习装置（40）通过从未标记的文本集中提取HMM对象名称统计信息，基于从标记的对象名称提取的HMM统计信息来推进共同训练学习，来扩展统计信息通过当前学习数据 文本集。 对象名识别器（50）通过维特比算法来确定形成输入文本的句子的语素的最佳HMM对象名统计信息路径来识别对象名称。

公开(公告)号：KR100327107B1

公开(公告)日：2002-03-06

申请号：KR1019990057300

申请日：1999-12-13

Applicant: 한국전자통신연구원

Inventor： 김현기 ,  노대식 ,  강현규 ,  박세영 ,  왕지현 ,  윤보현

IPC: G06F17/27

Abstract: 1. 청구범위에기재된발명이속한기술분야본 발명은문서관리시스템에서의정보재사용을위한에스지엠엘/엑스엠엘엔티티의저장/복원장치및 그방법에관한것임. 2. 발명이해결하려고하는기술적과제본 발명은, 다수의문서들에서참조될수 있는엔티티선언을중복하여저장하지않고공유하여사용할수 있도록하여보조기억장치의낭비를막고, SGML/XML 문서를보다빠르게저장하며, 데이터베이스에다양한형태로저장된엔티티들을엔티티유형에따라원래의엔티티들로복원할수 있도록한 문서관리시스템에서정보의재사용을위한 SGML/XML 엔티티의저장/복원장치및 그방법과상기방법을실현시키기위한프로그램을기록한컴퓨터로읽을수 있는기록매체를제공하고자함. 3. 발명의해결방법의요지본 발명은, 문서관리시스템에서의정보재사용을위한에스지엠엘(SGML)/엑스엠엘(XML) 엔티티의저장/복원장치에있어서, SGML/XML DTD(Document Type Definition) 및문서에대해파싱을수행하여오류의검증과파싱결과를주기억장치에트리형태로제공하기위한 SGML/XML 파싱수단; 상기 SGML/XML 파싱수단으로부터파싱결과를입력받아문서내에포함된엔티티참조부들을엔티티선언과연동하여저장하기위한문서저장수단; 상기 SGML/XML DTD의엔티티선언을저장하기위한엔티티저장수단; 사용자인터페이스로부터저장소식별자를입력받아저장소식별자에해당하는엔티티객체의데이터를반환하기위한엔티티관리수단; 및데이터베이스시스템과연동하기위해사용되는데이터베이스시스템인터페이싱수단을포함함. 4. 발명의중요한용도본 발명은문서관리시스템등에이용됨.