This paper describes an improved Continuous Dynamic Programming (CDP) for human gesture recognition. In order to cope with the variation in the amplitude of the input signal, a reference pattern set containing patterns with different mean amplitude is used instead of a single reference pattern. Furthermore, a new multiple search path determination strategy is developed in place of the traditional single one to deal with the variation in the speed of the input gesture. Also, a matching error controlled path determination method is newly introduced to reduce the influence of noise. The experimental results show this improved CDP can give not only stable identification results compared with the conventional one, but also the degree of the size and speed of the gesture to be distinguished.
The rating method and the method of successive categories, which have been developed in the field of psychophysics, are often utilized for rating the objects by use of the successive categories expressed by words. This paper points out several questions about the conventional methods and proposes a new method from the viewpoint that the rating conducts can be regarded as the logical judgement, taking into consideration the fuzziness in the meaning of words and the subjective difference of feeling. The proposed method is applied to a problem of rating the acoustic noise in time series and is experimentally shown to be useful for clarifying the criterion of judgement and the context effect.
In this paper, the entrance person reception system which uses the FG vision sensor is proposed as a person's behavior observation system. In this system, the observation is carried out by using the 3-D shape data obtained from the FG vision senor which is one of the range sensors. As an experimental examples, the FG vision sensor is installed at the top of the door. It is recognized whether it is a person who comes in, or a person who just passes, by detecting the behavior of the man who is here in front of the door. Even if there is a man of the wheelchair or a child, the height information of the object is not used so that the recognition may be carried out precisely. In this system, the recognition is done by the inclination information of the local plane.
This article discusses how to realize the process of converting the user's face into an avatar's face by preserving the expression. The objective of this work is to make this process reflect the user's subjective view about correspondence between the facial images of the user and the avatar. In this article, it is proposed to adapt the process to the user's view through interactions with the user. This is realized by an example-based approach, in which the correspondence between the facial images of the user and the avatar is estimated from its examples obtained from the interactions with the user. The process increasingly reflects the user's view with the number of examples.
We have developed a pen-based Japanese character input system for the blind person (particularly persons with acquired blindness). This system is composed of a personal computer and a control board with an electronic tablet. The blind person is able to get the screen information by using a voice synthesizer. We have investigated various problems that arise when the blind person edits the document by using this system and solved those problems. Furthermore this system is applied to an electronic mail system for the blind person. From the experimental results, we have confirmed that the proposed pen-based system makes it easy to input Japanese characters for the novice blind user without training.
The omnidirectional image sensor, called HyperOmniVision, can observe a 360 degree field of view and can transform an input image to a common camera image or a panoramic image easily. However, it has an intrinsical problem that the angular resolution of HyperOmniVision is lower than that of a conventional video camera. We propose a resolution improving method for HyperOmniVision, which fuses consecutive images obtained by turning motion of HyperOmniVision. We discuss the optimization of both optics and processing by simulation results.
Many methods of human motion measuring have been proposed and applied to various field, such as CG animation, gesture recognition, and so on. This paper proposes a method for posture estimation of human hand gesture. First, we extract human face and hand regions depended on their color and positions on stereo images taken by cameras, and estimate 3D positions on the world coordinate system by stereo matching. Then, we calculate occurence probabilities of motion model, and decide initial parameters related to human posture. Finally, we estimate the optimal posture by model matching, and predict the posture in the next frame.