Abstract:
The present invention relates to an extraction method for multi-touch feature information and a recognition method for multi-touch gestures using the multi-touch feature information, and more specifically, to the extraction method for multi-touch feature information and recognition method for multi-touch gestures using the multi-touch feature information, wherein: multi-touch feature information, which does not depend on the number of touch points, is extracted; and the accuracy in gesture recognition is improved by using the extracted multi-touch feature information.
Abstract:
PURPOSE: A portable terminal and touch recognition method thereof are provided to recognize gesture through plural touch signals by distinguishing the change of patterns for plural images. CONSTITUTION: A touch screen simultaneously receives plural touch signals. A camera unit(121) is arranged at the bottom of the touch screen. The camera unit successively receives plural images including the inputted plural touch signal during fixed time. A control unit(180) determines the change of touch patterns included in each image for each image. The control unit recognizes gestures for the touch signals according to determination results.
Abstract:
PURPOSE: A pose recognizing system using a joint model database is provided to improve a recognition speed and recognition rate. CONSTITUTION: A preprocessor(120) obtains a foreground image from an input image and changes the foreground into a distance converting image. A key pose recognizer(130) generates a joint model and a key pose eigenvectors and stores a pose library(110). The key pose recognizer extracts a key pose characteristic vector which is the most similar with a current image.
Abstract:
PURPOSE: An apparatus for improving gesture recognizing ratio in a mobile device is provided to generate integrated specification information from a plurality of input data, and recognize a gesture. CONSTITUTION: A gesture recognition unit(102) comprises a data acquisition unit(104), a feature extraction unit(106), and an integrated information generator(108). The data acquisition unit acquires a plurality of input data through cameras. The feature extraction unit performs pre-processing about a plurality of input data. The feature information extraction unit classifies the domain capable of gesture recognition. The integrated information generator mixes the extracted feature information to one integrated feature information.
Abstract:
PURPOSE: A recognition method of three dimensional human pose using a single camera is provided to estimate the photographing direction of human body by comparing angle histogram obtained from two dimensional pose image. CONSTITUTION: A three dimensional human body pose method for recognizing by a single camera is as follows. A target object having arbitrary three dimensional pose is photographed to obtain standard two dimensional pose images(S1000). A silhouette image of each standard two dimensional pose image is extracted, and the standard pose model is created based on the standard silhouette image. The standard pose model is stored in a standard pose model database. Standard pose models are created after repeating the first or third step, and stores in standard pose model database. The detection two dimensional pose image which takes a photograph of the detection object in the arbitrary direction by the single camera is gotten. The detected two dimensional pose image detection silhouette image is extracted which is based on the detected silhouette image and the video feature of the detection object. The standard pose model is obtained which has a closest similarity with the detection pose model.
Abstract:
본 발명은 영상내에 존재하는 인식대상(행위자)의 동작을 최적의 수치화된 모델을 통하여 표현함으로써 인식대상의 동작을 명확하게 구별할 수 있도록 한 가상의 격자형 평면을 이용한 동작 인식 방법에 관한 것으로, 스테레오 영상 획득 장치가 인식대상의 영상을 획득하는 제 1과정; 특징 추출 장치가, 입력되는 상기 영상의 정보를 기반으로 가상의 격자형 평면을 구성시키고, 그 구성시킨 가상의 격자형 평면을 이용하여 상기 입력영상의 동작에 대한 매 프레임마다의 특징벡터를 추출하는 제 2과정; 자세코드 시퀀스 생성기가, 입력되는 상기 매 프레임마다의 특징벡터를, 모델 동작 학습기를 통해 얻어진 동작공간에서 가장 가까운 거리를 갖는 자세코드로 변환하여 자세코드 시퀀스를 생성하는 제 3과정; 및 동작 인식기가, 입력되는 상기 매 프레임의 자세코드 시퀀스를, 학습에 의해 미리 형성된 모델 동작과 비교하여 상기 입력영상의 동작을 최종적으로 인식하는 제 4과정을 구비한다. 영상, 가상, 격자형 평면, 은닉마르코프 모델, 특징벡터, 동작인식