We address the problem of robust face identification in the presence of pose, lighting, and expression variation. Previous approaches to the problem have assumed similar models of variation for each individual, estima...
详细信息
ISBN:
(纸本)0818683449
We address the problem of robust face identification in the presence of pose, lighting, and expression variation. Previous approaches to the problem have assumed similar models of variation for each individual, estimated from pooled training data. We describe a method of updating a first order global estimate of identity by learning the class-specific con-elation between the estimate and the residual variation during a sequence. This is integrated with an optimal tracking scheme, in which identity variation is decoupled from pose, lighting and expression variation. The method results in robust tracking and a more stable Estimate of facial identity under changing conditions.
The capability of recognition is critical in learning but variation Of sensory input makes recognition a very challenging task. The current technology in computer vision and patternrecognition requires humans to coll...
详细信息
ISBN:
(纸本)0818683449
The capability of recognition is critical in learning but variation Of sensory input makes recognition a very challenging task. The current technology in computer vision and patternrecognition requires humans to collect images, stare images, segment images for computers and train computer recognition systems using these images. It is unlikely that such a manual labor process can meet the demands of many challenging recognition tasks that are critical for generating intelligent behavior, such as face recognition, object recognition and speech recognition. Our goal is to enable machines to learn directly from sensory input streams while interacting with the environment including human teachers. While doing so, the human teacher is not allowed to dictate the internal state value of the system. He or she can influence the system through only the system's sensors and effectors. Such a capability requires a fundamentally new way of addressing the learning problem, one that unifies learning and performance phases and requires a systematic self-organization capability. This paper concentrates on the state self-organization problem. We apply the method to autonomous face recognition.
We propose a method for learning models of human motion from a coarsely sampled set of examples. The models we synthesize may be used to generate plausible motions from a high level description consisting of start and...
详细信息
ISBN:
(纸本)0818683449
We propose a method for learning models of human motion from a coarsely sampled set of examples. The models we synthesize may be used to generate plausible motions from a high level description consisting of start and stop positions, style, mood, age, etc. In the field of computer vision, such models can be useful for human body motion tracking/estimation and gesture recognition. The models can also be used to generate arbitrary realistic human motion, and may be of help in trying to understand the mechanisms behind the perception of biological motion by the human visual system. Experimental results of the learning technique applied to reaching and drawing motions rare presented.
This paper proposes a human face indexing system on video data with face extraction, recognition, tracking and modeling function. The proposed system can extract, recognize and track the human faces regardless of thei...
详细信息
ISBN:
(纸本)0818683449
This paper proposes a human face indexing system on video data with face extraction, recognition, tracking and modeling function. The proposed system can extract, recognize and track the human faces regardless of their orientation and sizes. The time sections including tracked faces are recorded with their names as indices. These indices can be used to classify or retrieve the video data. If the face could not be recognized, the tracked facial images can be used for face modeling. We used subspace methods for the face extraction, recognition and tracking.
The proceedings contain 97 papers. The topics discussed include: beyond eigenfaces: probabilistic matching for face recognition;face verification based on morphological shape decomposition;a robust face identification...
ISBN:
(纸本)0818683449
The proceedings contain 97 papers. The topics discussed include: beyond eigenfaces: probabilistic matching for face recognition;face verification based on morphological shape decomposition;a robust face identification against lighting fluctuation for lock control;a multi-scale morphological method for human posture recognition;face indexing on video data - extraction, recognition, tracking and modeling;scale invariant face detection method using higher-order local autocorrelation features extracted from log-polar image;object oriented face detection using range and color information;detecting head pose from stereo image sequence for active face recognition;integration of eigentemplate and structure matching for automatic facial feature detection;and head pose determination from one image using a generic model.
It is essential for machine lip reading to process not only static images but also moving images. This paper proposes a method, which is named 'Optical-Snakes', for lip motion extraction from moving images usi...
详细信息
ISBN:
(纸本)0818683449
It is essential for machine lip reading to process not only static images but also moving images. This paper proposes a method, which is named 'Optical-Snakes', for lip motion extraction from moving images using the cooperation between the active contour models called Snakes and the optical flow. The effectiveness of the method is demonstrated by the exact extraction of a series of lip contours from time-varying images without any markers, and by the analysis of lip contours for machine lip reading. Some real products which characterize a speaker's pronunciation are made from the lip motion data by a photoforming system. It is demonstrated that the real products are better than the virtual products on a computer display for the evaluation of shapes. The application of lip motion to molding would lead to a new method for the analysis of lip motion and the training of pronunciation.
In this paper, we have proposed a physics-based facial model based on anatomical data and knowledge, which has a three dimensional structure for facial image analysis and synthesis. The skin and the skull models are c...
详细信息
ISBN:
(纸本)0818683449
In this paper, we have proposed a physics-based facial model based on anatomical data and knowledge, which has a three dimensional structure for facial image analysis and synthesis. The skin and the skull models are constructed from 3D-CT scanned data, and the facial muscles were placed between them just like as a real human face. In regards to the skin and muscle models, spring frames are used to simulate the elastic dynamics or a real facial skin and muscles. Facial Modification can be flexibly realized by two factors: the expansion and contraction of non-linear springs which simulate a skin and muscles, and the motion of the jaw part of the skull. Using our system, we can represent realistic facial animation and predict facial modifications under given conditions for muscles and the skull. Furthermore, our method is effective for analyzing relationship between modification on the facial surface and the motion of muscles and the skull.
In this paper, we propose a method to identify the human face using a 3D gray-scale image which combines a 3D image with a gray-scale image. The proposed method can identify facial images which face in various directi...
详细信息
ISBN:
(纸本)0818683449
In this paper, we propose a method to identify the human face using a 3D gray-scale image which combines a 3D image with a gray-scale image. The proposed method can identify facial images which face in various directions. First, 3D positions, where the eyes and the tip of the nose are located, are estimated in the acquired 3D gray-scale image. Next, the calibration of facial directions and brightness is performed. Due to facial images facing more to the right or the left, some parts of the calibrated data are lost by occlusion. Thus, we assume that a human face is almost symmetrical, and the side of the facial image without any loss of data is regarded as the region for identification. Feature vectors which reflect individually are extracted from this region. The identification procedure is conducted using the subspace method. In order to demonstrate the efficiency of this method, the experiment used 3D gray-scale images of 24 people.
A variety of techniques from statistics, signal processing, patternrecognition, machinelearning, and neural networks have been proposed to understand data by discovering useful categories. However, research in data ...
详细信息
The goal of pattern classification can be approached from two points of view: informative - where the classifier learns the class densities, or discriminative - where the focus is on learning the class boundaries with...
详细信息
暂无评论