Abstract:
1. 청구범위에 기재된 발명이 속한 기술분야 복수의 이미지 셋을 포함하는 비디오 시퀀스에 대한 감소된 비트를 갖는 에지 히스토그램 기술자의 비트 표현을 구성하는 방법 및 상기 에지 히스토그램 기술자의 부호화된 표현으로부터 효과적으로 추출된 정보를 이용하여 비디오 시퀀스를 검색하는 방법에 관한 것임. 2. 발명이 해결하고자 하는 기술적 과제 텍스쳐 비디오를 포함하는 대응 비디오 시퀀스를 검색하는 방법을 제공함. 3. 발명의 해결 방법의 요지 복수의 비디오 시퀀스를 표현하는 디지털 비디오 데이터 정보 - 상기 복수의 비디오 시퀀스 각각은 상기 디지털 비디오 데이터의 이미지 프레임 셋을 포함함 - 를 갖는 데이터베이스 구축 방법에 있어서, 상기 각 비디오 시퀀스의 상기 각 이미지 프레임을 L개(L은 양의 정수)의 부영상(sub-image) - 상기 각 부영상은 S x T개(S 및 T는 각각 양의 정수)의 이미지 블럭(image-block)으로 더욱 분할됨 - 으로 분할하는 제1단계; 상기 각 이미지 블럭에 대해 5개의 기준 에지(reference edge) - 상기 기준 에지는 4개의 방향성(directional) 에지 및 1개의 무방향성(non-directional) 에지를 포함함 - 중 어느 하나를 할당하여 상기 각 이미지 프레임에 대해 L개의 에지 히스토그램(edge histogram) - 상기 에지 히스토그램은 M개의 에지 히스토그램 빈(edge histogram bin)을 포함함 - 을 생성하는 제2 단계; 상기 각 에지 히스토그램에 포함된 상기 에지 히스토그램 빈을 S x T로 정규화하여 상기 각 이미지 프레임에 대해 M개의 정규화된 에지 히스토그램 빈을 생성하는 제3단계; 상기 각 이미지 프레임의 정규화된 에지 히스토그램 빈에 기초하여 상기 각 비디오 시퀀스에 대한 L개의 대표(representative) 에지 히스토그램을 생성하기 위해 상기 각 비디오 시퀀스에 대해 M개의 대표 에지 히스토그램 빈을 계산하는 제4단계; 및 상기 데이터베이스에 저장되는 상기 각 대표 에지 히스토그램에 대한 제2이미지 기술자로서 M개의 양자화 인덱스값을 생성하기 위해 상기 대표 에지 히스토그램 빈을 비선형적으로 양자화하는 제5단계를 포함함. 4. 발명의 중요한 용도 비디오 시퀀스 검색에 이용됨.
Abstract:
PURPOSE: A falling sense apparatus and a method thereof for notifying a falling through a communication unit are provided to actively prevent the accident due to the hurt from a fall by notifying an accident to a fall-down control server. CONSTITUTION: An accelerator sensor(110) senses acceleration and gradient information. A controller(130) decides upper parts by using the acceleration and gradient information. The controller analyzes consecutive patterns of the behaviors. If the hurt from a fall is determined, a communication unit(160) transmits the hurt from a fall generation information.
Abstract:
본 발명은 모션벡터를 이용한 행위 인식 장치 및 그 방법에 관한 것으로, 카메라를 통해 입력되는 영상에서 X축 평균 모션벡터 및 Y축 평균 모션벡터를 산출한 후 인식 테이블에 매핑하여 행위를 인식하기 위한, 모션벡터를 이용한 행위 인식 장치 및 그 방법을 제공하고자 한다. 이를 위하여, 본 발명은 모션벡터를 이용한 행위 인식 장치에 있어서, 평균 모션벡터에 상응하는 행위 형태 정보를 저장하고 있는 저장수단; 외부로부터 입력되는 영상을 수집하기 위한 영상 수집수단; 상기 영상 수집수단을 통해 수집한 영상에서 객체를 인식한 후 객체의 모션벡터를 산출하기 위한 모션벡터 산출수단;상기 모션벡터 산출수단에서 산출한 모션벡터에 대하여 평균 모션벡터를 산출하기 위한 평균 모션벡터 산출수단; 및 상기 평균 모션벡터 산출수단에서 산출한 평균 모션벡터로 상기 저장수단을 탐색하여 해당 행위를 인식하기 위한 행위 인식수단을 포함한다. 모션벡터, X축 평균 모션벡터, Y축 평균 모션벡터, 인식 테이블, 행위 인식
Abstract:
An apparatus for supporting taking medicine includes a medicine storage unit which has a plurality of medicine storage holes to store therein medicines to be taken by a user, a sensor controller which determines whether the medicine storage holes include medicines therein, and a medicine taking service unit which receives a preset schedule by synchronizing with a personal computer, checks whether a user has taken medicine according to a determination of the sensor controller and informs a user to take a medicine according to the preset schedule. Thus, users who should take various medicines on a regular basis for daily life may observe the medicine taking time.
Abstract:
A method and a system for daily life act recognition utilizing a sensor for recognizing a daily life act are provided to install the sensor in the furniture and floor in a house, thereby enabling a user to directly know the usage information of an article related to the daily life act. A wireless process management unit receives measured data from a plurality of sensors installed in a necessary location in a house in real time, recognizes a corresponding sensor which transmits the received measured data, and converts the received measured data into an elementary act corresponding to the recognized sensor. A buffer management unit stores the elementary action received from the wireless process management unit in an internal buffer. A daily life act recognition unit recognizes the daily life act of the aged in every time which is previously set up based on the elementary action stored in the internal buffer.
Abstract:
An apparatus and a method for detecting fall-down, and a system and a method for emergency aid using the same are provided to allow elderly persons to enjoy their safe life irrespective of a place and a time by using a portable terminal. An apparatus for detecting fall-down includes a storing unit, an angular velocity measuring unit, an acceleration measuring unit, an acceleration extracting unit(15), and a fall-down determining unit(16). The storing unit stores fall-down data vectors. The angular velocity measuring unit measures an angular velocity value. The acceleration measuring unit measures an acceleration value. The acceleration extracting unit extracts a kinetic acceleration value and a gravitational acceleration value by filtering the acceleration value measured by the acceleration measuring unit. The fall-down determining unit converts the angular velocity value which is measured by the angular velocity measuring unit, and the kinetic acceleration value and the gravitational acceleration value which are extracted by the acceleration extracting unit, into a fall-down data vector, and determines a user's fall-down by comparing the converted fall-down data vector with the fall-down data vector stored in the storing unit.
Abstract:
An apparatus and a method for managing feeding behavior using multiple sensors are provided to manage the eating behavior of a user efficiently by providing statistical information of an eating behavior in each period. An apparatus for managing feeding behavior using multiple sensors includes a learning information managing unit(31), a food information obtaining unit(32), an eating behavior information detecting unit and an eating behavior managing unit(37). The learning information managing unit manages the information on the kinds, components and state of food corresponding to the sensing information and image information. The eating behavior managing unit manages the eating behavior based on the kinds, components, state and eating amount of feed that are detected by the eating behavior information detecting unit.
Abstract:
An apparatus and a method for providing an ambient service on the basis of a ubiquitous environment, and a system using the same are provided to display three dimensionally processed image and application information for the old and to provide an interactive interface for enabling the old to output and to use the image easily. An apparatus for providing an ambient service includes an image database(20), a user identifier(21), a position & distance measuring unit(22), an image processor(23), and a controller(24). The image database stores image matched with proximity for each user. The user identifier identifies the user by using the RFID tag information from an RFID reader. The position & distance measuring unit measures the position of the user and the distance to the user by using ultrasonic waves emitted by an ultrasonic sensor. The image processor rearranges the image matched with the proximity on the basis of the position of the user. The controller sets the proximity according to the position of the user and the distance to the user, and controls the image processor for rearranging the image matched with the proximity on the basis of the position of the user.
Abstract:
An apparatus and a method for searching for ligands by using automatic extraction of a specific binding site are provided to design a drug with higher selectivity and to perform a protein structure based new drug design procedure, conveniently and rapidly. An apparatus for searching for ligands includes an input data processor(101), a binding site extractor(102), a user selecting input unit(103), manual comparing units(104,106), automatic comparing units(107,109), and a ligand searching unit(110). The input data processor processes inputted data, which is data on a protein. The binding site extractor extracts candidate binding sites, to which a drug is bound, from the data. The user selecting input unit allows a user to select a comparison scheme. The manual comparing units manually compare a binding site of a target protein with a similar binding site of a protein in a database. The automatic comparing units automatically compare a binding site of a target protein with a similar binding site of a protein in a database. The ligand searching unit performs virtual screening for the specific binding site from the manual comparing units and the automatic comparing units.
Abstract:
A method for clustering gene using gene expression profile is provided to obtain an accurate clustering result with low calculation complexity without directly designating complicated and detrimental input variations determining quality of a clustering result by clustering a time series gene expression profile data set into a gene group having the similar function. A method for clustering gene using gene expression profile comprises the steps of: (a) calculating a parameter value of each orders of a linear regression model when the gene expression profile is inputted; (b) calculating an F-distribution based probability regarding each of the parameter value calculated linear regression model; and (c) after selecting the linear regression model having a minimum value of the calculated F-distribution based probability, clustering the gene in accordance with the order of the selected linear regression model, wherein the gene is clustered by an increase function or a decrease function according to a sign of the corresponding parameter value when the order of the finally selected linear regression model is 1st(linear) or at least 3rd. or is clustered by a concave function or a convex function according to the sign of the corresponding parameter value when the order of the finally selected linear regression model is 2nd(parabola).