-
公开(公告)号：KR101079743B1
公开(公告)日：2011-11-03
申请号：KR1020100000840
申请日：2010-01-06
Applicant: 한국과학기술원
IPC: G10K15/00
Abstract: 본발명은멜로디라인결정방법에관한것으로, 해결하고자하는기술적과제는멜로디라인이가지는물리적특성에기반하여효율적으로멜로디라인을결정할수 있는방법을제공하는데있다. 이를위해본 발명에따른멜로디라인결정방법은, 각프레임에서 N개의멜로디피치후보들과상기피치후보들이가지는비중에관한정보를입력받는제 1단계와, 오디오신호의단위를헤르쯔(Hz)에서센트(Cent)로변환하는제 2단계와, 멜로디라인의시작프레임을설정하는제 3단계와, 상기제 3단계를통하여설정된시작프레임에서상기제 1단계에서입력받은 N개의멜로디피치후보들을선택하고이들을비중이큰 순으로정렬하는제 4단계와, 이웃한두 멜로디피치후보가기설정된멜로디라인기준을만족하는지판정하는제 5단계와, 멜로디라인연결을통해각 멜로디구간마다 N개의멜로디라인후보를선택하는제 6단계와, 상기제 6단계를통하여선택된 N개의멜로디라인후보중에서최적의멜로디라인을선택하는제 7단계와, 상기제 7단계에서선택된최적의멜로디라인을스무드화(Smoothing) 하는제 8단계를포함한다.
-
公开(公告)号：KR1020110085090A
公开(公告)日：2011-07-27
申请号：KR1020100004696
申请日：2010-01-19
Applicant: 한국과학기술원
IPC: G10L25/90 , G10L19/02 , G10L19/022 , H03M7/30
Abstract: PURPOSE: A frequency analyzing method using an analyzing window is provided to supply more reliable and exact frequency information analyzing method to exact accurate melody extraction and music content information in a audio signal having various sound. CONSTITUTION: An inputted audio signal is resampled(S10). An audio signal inputted based on a time area is converted into a signal based on a frequency area(S20). A peak value of a amplitude in and a frequency value of a location expressing a peak value in spectrum of each obtained frame are extracted(S30). Based on the extracted values, a range that a melody pitch of each frame exists is reset(S40). Dynamic change information of the melody pitch is extracted with a method that grasps an autocorrelation coefficient between frames(S50).
Abstract translation: 目的：提供使用分析窗口的频率分析方法，以提供更可靠和精确的频率信息分析方法，以便在具有各种声音的音频信号中精确的旋律提取和音乐内容信息。 构成：对输入的音频信号进行再采样（S10）。 基于时间区域输入的音频信号被转换为基于频率区域的信号（S20）。 提取在每个获得的帧中表示频谱中的峰值的位置的振幅和频率值的峰值（S30）。 基于提取的值，复位每个帧的旋律间距的范围（S40）。 通过掌握帧之间的自相关系数的方法来提取旋律音调的动态变化信息（S50）。
-
公开(公告)号：KR1020110017760A
公开(公告)日：2011-02-22
申请号：KR1020090075392
申请日：2009-08-14
Applicant: 한국과학기술원
CPC classification number: G06Q50/10 , G06F17/30017 , G06F17/30861 , G06K9/00 , G06Q20/12 , G10L15/00
Abstract: PURPOSE: A system for extracting a descriptor of a multimedia contents is provided to extract information of a multimedia contents and express the information through the descriptor. CONSTITUTION: A web server(200) offers a web site to a user terminal(100). A multimedia dictionary service server(300) analyzes the multimedia contents of the user terminal. The multimedia dictionary service server extracts the descriptor from the multimedia contents information. The server additionally inputs the descriptor to the corresponding multimedia contents. The service server searches the corresponding multimedia contents using the descriptor. The service server offers the searched multimedia contents to the user terminal.
Abstract translation: 目的：提供一种用于提取多媒体内容的描述符的系统，以提取多媒体内容的信息，并通过描述符表达信息。 规定：网络服务器（200）向用户终端（100）提供网站。 多媒体词典服务器（300）分析用户终端的多媒体内容。 多媒体词典服务服务器从多媒体内容信息中提取描述符。 服务器另外将描述符输入到相应的多媒体内容。 服务服务器使用描述符搜索相应的多媒体内容。 服务服务器将搜索到的多媒体内容提供给用户终端。
-
44.发明公开
公开(公告)号：KR1020110011969A
公开(公告)日：2011-02-09
申请号：KR1020090069471
申请日：2009-07-29
Applicant: 한국과학기술원
Abstract: PURPOSE: A method for establishing a model which can recognize feelings in a voice through a loss function and a maximum margin technique based on WTM(Watson-Tellegen Emotional Model) is provided to remarkably increase feeling recognition performance included in a voice. CONSTITUTION: A difference between each emotional feelings is figured by suing geometric distance between emotion groups of WTM(310). Based on set values in the first step, a value of a loss function is obtained(330). Based on a loss function, a parameter of each speech emotion module through a max-margin with margin scaling method is obtained(340).
Abstract translation: 目的：提供一种能够通过基于WTM（Watson-Tellegen情感模型）的损失函数和最大余量技术来识别语音感知的模型的方法，以显着增加语音中包含的感觉识别性能。 构成：通过起诉WTM情感组之间的几何距离（310）来计算每种情感感受之间的差异。 基于第一步骤中的设定值，获得损失函数的值（330）。 基于损失函数，获得每个语音情感模块通过最大余量与边缘缩放方法的参数（340）。
-
公开(公告)号：KR1020100034153A
公开(公告)日：2010-04-01
申请号：KR1020080093141
申请日：2008-09-23
Applicant: 한국과학기술원
IPC: G06F17/30
CPC classification number: G06F17/30831 , G06F17/30784 , G06F17/30802
Abstract: PURPOSE: A method and an apparatus for searching an image based on contents are provided to extract a feature vector from a searched specific area after searching a specific area of the image based on contents of an image. CONSTITUTION: An area search module(140) extracts a specific area of an area darker than the surround. A fingerprint extraction module(150) extracts a fingerprint vector of the specific area. A similarity calculation module(160) compares the fingerprint vector with a fingerprint vector of another image stored in an image information DB by an area.
Abstract translation: 目的：提供一种基于内容搜索图像的方法和装置，用于基于图像的内容搜索图像的特定区域之后，从搜索到的特定区域中提取特征向量。 构成：区域搜索模块（140）提取比环绕图像暗的区域的特定区域。 指纹提取模块（150）提取特定区域的指纹矢量。 相似度计算模块（160）将指纹矢量与存储在图像信息DB中的另一图像的指纹矢量区域进行比较。
-
Patent Agency Ranking
公开(公告)号：KR102259207B1
公开(公告)日：2021-05-31
申请号：KR1020130169041
申请日：2013-12-31
Abstract: 자동태깅시스템및 그방법이개시된다. 여기서, 자동태깅시스템은정지영상을입력받는입력부, 상기정지영상을해석하여객체의시각적인특징을포함하는물리적인정보를추론하는물리적정보추출부, 상기정지영상을해석하여상기정지영상의속성에해당하는의미론적인정보를추론하는의미론적정보추출부, 그리고상기정지영상의메타데이터, 상기물리적인정보및 상기의미론적인정보를통합하여상기정지영상에태깅하는자동태깅부를포함한다.
-
-
公开(公告)号：KR1020150079064A
公开(公告)日：2015-07-08
申请号：KR1020130169041
申请日：2013-12-31
CPC classification number: G06K9/00718 , G06F17/30268 , G06K9/46 , G06K9/6202 , G06T7/194 , G06T7/246 , G06T2207/20036
Abstract: 자동태깅시스템및 그방법이개시된다. 여기서, 자동태깅시스템은정지영상을입력받는입력부, 상기정지영상을해석하여객체의시각적인특징을포함하는물리적인정보를추론하는물리적정보추출부, 상기정지영상을해석하여상기정지영상의속성에해당하는의미론적인정보를추론하는의미론적정보추출부, 그리고상기정지영상의메타데이터, 상기물리적인정보및 상기의미론적인정보를통합하여상기정지영상에태깅하는자동태깅부를포함한다.
Abstract translation: 公开了一种自动标签系统及其方法。 本发明的自动标签系统包括：用于接收静止图像的输入单元; 物理信息提取单元，用于通过解释静止图像来推断包括对象的视觉特征的物理信息; 语义信息推理单元，用于通过解释静止图像来推断与静止图像的属性相对应的语义信息; 以及自动标记单元，用于通过集成静止图像的元数据，物理信息和语义信息来标记静止图像。
-
公开(公告)号：KR101403599B1
公开(公告)日：2014-06-03
申请号：KR1020130029398
申请日：2013-03-19
Applicant: 한국과학기술원
Abstract: An embodiment of the present invention, which relates to a method to generate a robot dance motion expression, provides a method to generate a robot dance motion expression including: an information extraction step of extracting a first music information from a music signal; an information prediction step of predicting a second music information so as to express a dance motion with the first music information; and an information selection step of selecting a dance motion-associated information suitable for the second music information from a motion database.
Abstract translation: 本发明涉及产生机器人舞蹈运动表达的方法的实施例提供了一种生成机器人舞蹈运动表达的方法，包括：信息提取步骤，从音乐信号中提取第一音乐信息; 信息预测步骤，用于预测第二音乐信息以便用第一音乐信息表达舞蹈动作; 以及信息选择步骤，从运动数据库中选择适合于第二音乐信息的舞蹈运动相关信息。
-
公开(公告)号：KR101348904B1
公开(公告)日：2014-01-09
申请号：KR1020120006668
申请日：2012-01-20
Applicant: 한국과학기술원
IPC: G06T7/00
CPC classification number: G06T7/0081 , G06T7/11 , G06T7/162
Abstract: 본 발명의 일 실시예에 따른 이미지 분할 방법은 입력 이미지를 슈퍼픽셀들로 나누는 제1 단계; 상기 제1 단계를 통해 얻어진 슈퍼픽셀들 중 인접하는 2 이상의 복수의 슈퍼픽셀을 특정 조건을 고려한 연결을 통해 하이퍼그래프를 구축하는 제2 단계; 및 상기 하이퍼그래프의 각 에지의 특징 벡터를 추출하여 조인트 특징 맵을 구성하고, 고차 상관 클러스터링을 통해, 구축된 상기 하이퍼그래프를 분할하는 제3 단계를 포함한다.
Abstract translation: 本文公开了一种基于使用高阶相关聚类的图像分割来处理图像的方法。 在根据本发明的实施例的图像分割方法中，将输入图像分割成超像素。 通过将超像素之中的两个或更多个相邻的超像素彼此连接来构造超图。 通过从超图的相应边缘提取特征向量来创建联合特征图，并且考虑到特定约束，基于高阶相关聚类划分超图。
-
-
-
-
-
-
-
-
-
Search Results
73
records
(took 0.024 seconds)
