-
公开(公告)号：KR101034117B1
公开(公告)日：2011-05-13
申请号：KR1020090109727
申请日：2009-11-13
Applicant: 성균관대학교산학협력단
IPC: G06T7/40
CPC classification number: G06K9/4609 , G06K9/3241 , G06T5/50 , G06T2207/20224
Abstract: PURPOSE: An object recognition method and a device thereof using an interest area definition and an outline image in an image. CONSTITUTION: A first preprocessor(4) performs a preprocessing of an image for recognizing a rotated image. A second preprocessor(5) performs a grey image conversion of an image and a normalization of a size of the outline image. A recognition candidate reducing unit(7) reduces the number of recognizing candidate through a comparison between outline information of an object and an outline information of an image. A recognition unit(8) recognizes an object of the recognizing candidate. A recognizing result confirmation unit(9) checks a recognizing result of the recognized object.
Abstract translation: 目的：一种在图像中使用兴趣区域定义和轮廓图像的对象识别方法及其装置。 构成：第一预处理器（4）执行用于识别旋转图像的图像的预处理。 第二预处理器（5）执行图像的灰度图像转换和轮廓图像的尺寸的归一化。 识别候选减少单元（7）通过对象的轮廓信息和图像的轮廓信息之间的比较来减少识别候选者的数量。 识别单元（8）识别识别候选的对象。 识别结果确认单元（9）检查识别对象的识别结果。
-
62.发明公开
公开(公告)号：KR1020090124629A
公开(公告)日：2009-12-03
申请号：KR1020080050947
申请日：2008-05-30
Applicant: 성균관대학교산학협력단
IPC: H04N21/236 , H04N21/25
CPC classification number: G06Q30/02
Abstract: PURPOSE: An interest information recommend system and a method thereof for supplying interest information to a user are provided to detect a relation rule by measuring the degree of interest and reproduction time of a user request contents. CONSTITUTION: A plurality of user terminals(10) supplies contents transmitted from an outer server. An inter contents relation in the user interest information deduction server offers actual reproduction time information based on maximum reproduction time of the contents. In case the contents request is received from a random terminal, the contents providing server offers the contents related to the requested contents according to the inferred associative relationship to the terminal.
Abstract translation: 目的：提供兴趣信息推荐系统及其向用户提供兴趣信息的方法，以通过测量用户请求内容的兴趣程度和再现时间来检测关系规则。 构成：多个用户终端（10）提供从外部服务器发送的内容。 用户兴趣信息扣除服务器中的内容关系基于内容的最大再现时间提供实际的再现时间信息。 在从随机终端接收到内容请求的情况下，内容提供服务器根据与终端的推断关联关系提供与所请求内容相关的内容。
-
公开(公告)号：KR1020090113617A
公开(公告)日：2009-11-02
申请号：KR1020080039426
申请日：2008-04-28
Applicant: 성균관대학교산학협력단
CPC classification number: G06T5/50 , G06K9/4614 , G06K9/6256
Abstract: PURPOSE: An image authentication method, a device thereof, and a computer readable record-medium recorded with a program for executing the method are provided to recognize an inputted image using entropy value of the difference image about multiple registered images. CONSTITUTION: A pre-processor(104) extracts a target image for a selection from an input image. A difference image acquisition unit(106) reads the average reference image about the target image from the save area. The difference image acquisition unit compares the average reference image with the target image. The difference image acquisition unit obtains the difference image following the comparison result. A calculating unit(108) produces the entropy value of the obtained difference image. A determining unit(110) compares the calculated entropy value and the designated selection threshold value.
Abstract translation: 目的：提供一种图像认证方法，其装置和记录有用于执行该方法的程序的计算机可读记录介质，以使用关于多个注册图像的差分图像的熵值来识别输入图像。 构成：预处理器（104）从输入图像中提取用于选择的目标图像。 差分图像获取单元（106）从保存区域读取关于目标图像的平均参考图像。 差分图像获取单元将平均参考图像与目标图像进行比较。 差分图像获取单元获得跟随比较结果的差分图像。 计算单元（108）产生所获得的差分图像的熵值。 确定单元（110）比较计算出的熵值和指定的选择阈值。
-
公开(公告)号：KR100902322B1
公开(公告)日：2009-06-12
申请号：KR1020070140071
申请日：2007-12-28
Applicant: 성균관대학교산학협력단
IPC: G06F17/30
CPC classification number: G06F17/3064 , G06F2216/03
Abstract: An apparatus for offering an acoustic/visual hybrid keyword based on a cooperative agent and a method therefor are provided to investigate the associated rule of keywords captured on an acoustic web and a visual web. Keyword capturing units(31,32) capture keywords which a user uses an acoustic web and a visual web, and constructs the captured keywords through the transaction. A data mining agent(33) investigates a keyword association rule for each of two web transactions. A rule integrating agent(34) integrates the two kinds of association rules. Keyword providing units(35,36) receive keywords which the user inputs in the acoustic web and visual web, and then provides the related keywords based on the association rules.
Abstract translation: 提供了一种用于提供基于协作代理及其方法的声/视混合关键词的装置，以调查在声学网和视觉网上捕获的关键关联规则。 关键词捕获单元（31,32）捕获用户使用声音网页和可视化网页的关键字，并通过交易构建捕获的关键字。 数据挖掘代理（33）调查两个Web事务中的每一个的关键字关联规则。 规则集成代理（34）集成了两种关联规则。 关键词提供单元（35,36）接收用户在声学网络和可视网页中输入的关键字，然后基于关联规则提供相关的关键字。
-
公开(公告)号：KR1020070119156A
公开(公告)日：2007-12-20
申请号：KR1020060053395
申请日：2006-06-14
Applicant: 성균관대학교산학협력단
CPC classification number: G06K9/00134 , G06F21/32 , G06K9/00221 , G06K9/4633
Abstract: A biometric recognition system using a teeth image, a method thereof, and a recording medium storing the same are provided to improve a recognition rate by recognizing a face or a voice of the user, and recognizing a structural feature of the teeth image different from each user at the same time. An image input part(101) receives a biometric image, which includes the teeth image, of a speaker. An image preprocessor(102) extracts a plurality of feature parameters by detecting a teeth part from the biometric image, normalizing the teeth part to a normal size, and extracting outline information from the normalized image. A model generator(103) generates and stores a biometric model to a database(104) by using the extracted feature parameters. A recognizer(105) recognizes a similar object by pattern matching between the extracted feature parameters and the biometric models stored in the database. An authenticator(106) discriminates identity between biometric information of the recognized similar object and the speaker. A voice input part(201) receives the voice of the speaker. A voice preprocessor(202) extracts a plurality of feature parameters from the received voice.
Abstract translation: 提供了一种使用牙齿图像的生物识别系统，其方法和存储该图像的记录介质，以通过识别用户的脸部或声音来提高识别率，并且识别与每个不同的牙齿图像的结构特征 用户同时进行。 图像输入部（101）接收扬声器的包含齿图像的生物体图像。 图像预处理器（102）通过检测来自生物图像的牙齿部分来提取多个特征参数，将牙齿部分归一化为正常大小，以及从归一化图像提取轮廓信息。 模型生成器（103）通过使用所提取的特征参数来生成并存储生物特征模型到数据库（104）。 识别器（105）通过提取的特征参数和存储在数据库中的生物特征模型之间的模式匹配来识别相似的对象。 鉴别器（106）识别识别的相似对象的生物特征信息与说话者之间的身份。 声音输入部（201）接收扬声器的声音。 语音预处理器（202）从接收的语音中提取多个特征参数。
-
Patent Agency Ranking
公开(公告)号：KR102254740B1
公开(公告)日：2021-05-24
申请号：KR1020190128038
申请日：2019-10-15
Applicant: 성균관대학교산학협력단
IPC: A61B5/00 , G16H50/20 , A61B5/0205 , G06N3/08 , G06N3/04
Abstract: 본발명의스마트신호인식기는사용자를센싱하여센싱정보를생성하는센서;와상기센서에서전송된상기센싱정보를멀티레이블딥러닝학습하도록제어하는프로세서;를포함하고, 상기프로세서는, 센싱정보에대해사용자선택형복수의전처리과정을수행하고, 상기복수의전처리과정각각에의해생성된 3차원데이터를저장하고, 상기복수의전처리과정각각에의해생성된 3차원데이터를 2차원데이터로압축하고, 압축된상기 2차원데이터에대해, 멀티레이블딥러닝학습을수행하고, 상기멀티레이블딥러닝학습을통해, 스마트신호인식모델을생성하는것을포함한다.
-
公开(公告)号：KR1020210044625A
公开(公告)日：2021-04-23
申请号：KR1020190128038
申请日：2019-10-15
Applicant: 성균관대학교산학협력단
IPC: A61B5/00 , G16H50/20 , A61B5/0205 , G06N3/08 , G06N3/04
Abstract: 본발명의스마트신호인식기는사용자를센싱하여센싱정보를생성하는센서;와상기센서에서전송된상기센싱정보를멀티레이블딥러닝학습하도록제어하는프로세서;를포함하고, 상기프로세서는, 센싱정보에대해사용자선택형복수의전처리과정을수행하고, 상기복수의전처리과정각각에의해생성된 3차원데이터를저장하고, 상기복수의전처리과정각각에의해생성된 3차원데이터를 2차원데이터로압축하고, 압축된상기 2차원데이터에대해, 멀티레이블딥러닝학습을수행하고, 상기멀티레이블딥러닝학습을통해, 스마트신호인식모델을생성하는것을포함한다.
-
公开(公告)号：KR102241483B1
公开(公告)日：2021-04-19
申请号：KR1020180082956
申请日：2018-07-17
Applicant: 성균관대학교산학협력단
IPC: A61B5/00
Abstract: 떨림이란병리적또는생리적으로손, 머리, 기타신체부위에서발생된다. 병리적인떨림은몸의일부또는전체가규칙적인리듬을가지고떨리는증세를말한다. 이상운동질환중에서가장흔한증상이며발생하는원인에따라여러가지종류가있다. 떨림의의학적용어인진전의병태생리는아직정확하게밝혀져있지않으나, 중추신경계에진전을발생시키는근원지가있어진전이나타난다는가설과소뇌기능의이상으로인하여진전이발생한다는가설이대표적이다. 또한, 떨림의횟수와발생시기에따라다양한질환의전조현상또는합병증세로나타나는경우가있다. 진전증세를동반하는대표적질환으로는파킨슨병이있다. 파킨슨병의진전증세는환자가쉬고있을때 나타나며자발적인운동을하는동안에는떨림이감소하는안정시 진전양상을보인다. 안정시 진전이란손을무릎위나의자등 가만히얹어놓고있는상태에서떨림이심하게발생하고물건을들거나잡고있으면떨림이감소하는양상의진전을말한다. 병리적증세에의한떨림외에도대부분의사람들은일정한패턴의떨림을가지고있다. 이러한떨림의패턴은사용자를분류하는특징으로사용할수 있으며, 본발명에서는떨림패턴을이용한사용자인식과더불어사용자별병적징후를예측하는방법에대한발명이다.
-
-
-
-
-
-
-
-
Search Results
129
records
(took 0.021 seconds)
