会员体验
专利管家(专利管理)
工作空间(专利管理)
风险监控(情报监控)
数据分析(专利分析)
侵权分析(诉讼无效)
联系我们
交流群
官方交流:
QQ群: 891211   
微信请扫码    >>>
现在联系顾问~
热词
    • 3. 发明公开
    • 연속 음성인식 모델 구축방법
    • 连续语音识别模型构建方法
    • KR1020060092683A
    • 2006-08-23
    • KR1020050013760
    • 2005-02-18
    • 엘지전자 주식회사
    • 신원호
    • G10L15/08
    • 본 발명은 연속 음성인식 모델 구축방법에 관한 것으로, 특정 음성인식 관련 분야별로 언어모델 확률값을 대표할 수 있는 대표 단어를 설정하고, 대표단어의 그 언어모델 확률값을 나머지 단어에 대하여 적용함으로써, 언어모델 구축에 필요한 코퍼스 생성시간 및 작업량을 감소시키도록 한 것이다. 이를 위하여 본 발명은, 사용자가 음성인식 적용분야를 선택하는 과정과; 상기 음성인식 적용분야에 해당되는 다양한 패턴의 문장들을 문법으로 표현하는 과정과; 상기 음성인식 적용분야에 해당되는 대표단어를 설정하는 과정과; 상기 대표단어 및 문법을 이용하여 다양한 텍스트 코퍼스를 생성하는 과정과; 상기 다양한 텍스트 코퍼스를 이용하여 언어모델을 생성하는 과정과; 상기 대표단어와 동일분야에 해당되는 단어를 추가사전 리스트에 맵핑하는 과정으로 이루어진다.
    • 5. 发明公开
    • 소음환경에서의 끝점 검출 방법
    • 检测噪声环境中端点的方法
    • KR1020020095502A
    • 2002-12-27
    • KR1020010033446
    • 2001-06-14
    • 엘지전자 주식회사
    • 신원호
    • G10L15/20
    • PURPOSE: A method for detecting an end point in a noise environment is provided to detect voice with little calculation amount, prevent end point detecting performance from lowering in the case that a speaker speaks during an automatic response service. CONSTITUTION: A method for detecting an end point in a noise environment includes the steps of extracting predetermined parameters required for detecting an end point and voice recognition(S300), confirming whether a current frame is voice section or non-voice section(S310), updating parameter data of the voice section if the current frame is confirmed as the voice section(S320), updating parameter data of the non-voice section if the current frame is confirmed as the non-voice section(S330), extracting a start point and the end point of the voice through the updated parameter data(S340), confirming whether the extraction of the end point is successfully terminated(S350), and detecting relatively accurate voice in a noise environment similar to the voice of a speaker(S360).
    • 目的:提供一种用于检测噪声环境中的终点的方法,用于以较少的计算量来检测语音,防止在自动响应服务期间说话者说话的情况下终点检测性能降低。 构成:用于检测噪声环境中的终点的方法包括以下步骤:提取检测终点和语音识别所需的预定参数(S300),确认当前帧是语音部分还是非语音部分(S310), 如果当前帧被确认为语音部分,则更新语音部分的参数数据(S320),如果当前帧被确认为非语音部分,则更新非语音部分的参数数据(S330),提取起始点 和通过更新的参数数据的语音终点(S340),确认结束点的提取是否成功终止(S350),并且在类似于扬声器的声音的噪声环境中检测相对准确的语音(S360) 。
    • 6. 发明公开
    • 반향 제거 장치
    • ECHO CANCELER DEVICE
    • KR1020140006195A
    • 2014-01-16
    • KR1020120069175
    • 2012-06-27
    • 엘지전자 주식회사
    • 조성문전혜정신원호
    • H04R3/02
    • H04R3/02G10L2021/02082H04M9/082H04R2499/11
    • In accordance with an embodiment of the present invention, an echo removal device removes echo signals by being mounted on a first terminal. The echo removal device includes a first adaptive filter generating a first standard signal through a voice signal of a far-end talker received from a second terminal and removing a first echo signal corresponding to the voice signal received from the second terminal; and a second adaptive filter generating a second standard signal through a voice signal of a near-end talker inputted to a first terminal and removing a second echo signal corresponding to the voice signal in which the first terminal receives. [Reference numerals] (151) First adaptive filter; (153) Second adaptive filter
    • 根据本发明的实施例,回波去除装置通过安装在第一端子上来去除回波信号。 回波去除装置包括:第一自适应滤波器,通过从第二终端接收的远端讲话者的语音信号产生第一标准信号,并移除与从第二终端接收到的语音信号相对应的第一回波信号; 以及第二自适应滤波器,通过输入到第一终端的近端讲话者的语音信号产生第二标准信号,并且去除与第一终端接收的语音信号对应的第二回波信号。 (附图标记)(151)第一自适应滤波器; (153)第二自适应滤波器
    • 8. 发明公开
    • 음성인식을 이용한 언어 번역 방법 및 장치
    • 使用语音识别翻译语言的方法和装置
    • KR1020100019596A
    • 2010-02-19
    • KR1020080078247
    • 2008-08-11
    • 엘지전자 주식회사
    • 김유진신원호
    • G10L15/08H04B1/40
    • G06F17/28G06F17/289G10L15/22
    • PURPOSE: A translating method using voice recognition and a device using the same are provided to obtain translation results of the language which a user speaks. CONSTITUTION: A voice recognition unit(103) enables the voice recognition of inputted voice. The voice recognition unit obtains one or more recognition candidates corresponding to the inputted voice. According to the number of the obtained recognition candidates, a controller(119) provides one among user interfaces. The controller receives a selection signal of one among the obtained user interface through the provided user interface. The controller outputs translation data consisting of a second language corresponding to the recognition candidate.
    • 目的:提供使用语音识别的翻译方法和使用其的设备,以获得用户所说的语言的翻译结果。 构成:语音识别单元(103)能够进行输入语音的语音识别。 语音识别单元获得与所输入的语音相对应的一个或多个识别候选。 根据所获得的识别候选的数量,控制器(119)在用户界面中提供一个。 控制器通过提供的用户界面接收获得的用户界面之一的选择信号。 控制器输出由与识别候选对应的第二语言组成的翻译数据。
    • 9. 发明公开
    • 텍스트/음성변환을 위한 음소 경계 조정방법
    • 用于调整文本/语音转换的方位边界的方法
    • KR1020030083903A
    • 2003-11-01
    • KR1020020022300
    • 2002-04-23
    • 엘지전자 주식회사
    • 신원호
    • G10L13/08
    • PURPOSE: A method for adjusting a phoneme boundary for text/speech conversion is provided to correct an error in a phoneme boundary obtained from a voice recognizing device to offer more accurate phoneme boundary information. CONSTITUTION: Phoneme information recognized by a voice recognizing device and boundary values of the phoneme information are sequentially inputted(S1), to separately detect 'voiceless consonant and vowel' and 'vowel+voiceless consonant'(S2,S3). A position where a smoothed energy variation rate represents a maximum value is controlled to be a boundary of 'consonant+vowel' or 'consonant+consonant'(S7). A position where the slope of smoothed ZCR represents a maximum value is controlled to be a boundary of 'vowel+consonant'.
    • 目的:提供一种用于调整文本/语音转换的音素边界的方法,以校正从语音识别装置获得的音素边界中的错误,以提供更准确的音素边界信息。 构成:依次输入语音识别装置识别的音素信息和音素信息的边界值(S1),分别检测“无声辅音和元音”和“元音+无声辅音”(S2,S3)。 将平滑的能量变化率表示为最大值的位置被控制为“辅音+元音”或“辅音+辅音”的边界(S7)。 平滑ZCR的斜率表示最大值的位置被控制为“元音+辅音”的边界。
    • 10. 发明公开
    • 학습된 모델을 이용한 상태 클러스터링 방법
    • 使用学习模型进行状态聚类的方法
    • KR1020020095503A
    • 2002-12-27
    • KR1020010033447
    • 2001-06-14
    • 엘지전자 주식회사
    • 신원호
    • G10L15/183G10L15/14
    • PURPOSE: A method for state clustering by using a learned model is provided to reduce calculation amount by a lot of data by executing a state clustering with the learned model. CONSTITUTION: A method for state clustering by using a learned model includes the steps of inputting a new word(s301), decomposing the input word into a plurality of triphones and confirming whether the triphones are learned triphones(s302), applying an acoustic question by replacing with the learned triphone if the learned triphone exists in the triphones(s303), selecting the most adjacent triphone in the learned triphones(s304), selecting the optimum question in the questions of a question set(s305), dividing a cluster(s306), granting weight by obtaining frequency of each triphone(s307), and judging whether the divided cluster is larger than a critical value(s308).
    • 目的:提供一种通过使用学习模型进行状态聚类的方法,通过使用学习模型执行状态聚类来减少大量数据的计算量。 构成:通过使用学习模型的状态聚类的方法包括输入新词(s301)的步骤,将输入词分解为多个三音节并确认三电话是否是学习三电话(s302),应用声学问题由 如果学习的三通电话存在于三通电话(s303)中,则替换学习的三通电话(s303),选择所学习的三通电话中最相邻的三星电话(s304),在问题集(s305)的问题中选择最佳问题,划分集群(s306) ),通过获得每个三通电话的频率来授权(s307),以及判断分割的集群是否大于临界值(s308)。