The use of human motions for the interaction between humans and computers is becoming an attractive alternative to verbal media, especially through the visual interpretation of the human body motion. In particular, hand gestures are used as non-verbal media for the humans to communicate with machines that pertain to the use of the human gestures to interact with them. This paper introduces a 3D motion measurement of the human upper body for the purpose of the gesture recognition, which is based on the probability distribution of arm trajectories. In this study, by examining the characteristics of the arm trajectories given by a signer, motion features are selected and classified by using a fuzzy technique. Experimental results show that the use of the features extracted from arm trajectories effectively works on the recognition of dynamic gestures of a human, and gives a good performance to classify various gesture patterns.
This paper proposes a new algorithm of generating command pulses for numerical controllers. Differently from the presently used NC pulse generating method, pulse trains with automatically changed frequency following reference displacement commands can be generated using the proposed algorithm, and the generated pulse command can be also modified in real time by introducing sensor signals into the controller feedback loop. Effectiveness of the proposed algorithm is demonstrated by simulation results, and a method of employing a feedforward controller to improve the controller performance is also discussed in this paper.
This paper proposes a decentralized model predictive control method for solving an optimal consensus problem, where a system consists of networked multiple subsystems and the states of all the subsystems converge to a common point. The problem is formulated as a convex optimization problem involving linear matrix inequalities, and then is solved by using an incremental subgradient method based on primal decomposition. In the proposed scheme, the state feedback matrix for each subsystem is computed at each time in a decentralized way. It is shown that the states of all the subsystems asymptotically converge by the proposed method if the optimization problem is feasible at the initial time. A numerical example is given to show the effectiveness of the proposed method.
This paper details a network-controlled measurement system for use in fisheries engineering. The target strength (TS) of fish is important in order to convert acoustic integration values obtained during acoustic surveys into estimates of fish abundance. The target strength pattern is measured with the combination of the rotation system for the aspect of the sample and the echo data acquisition system using the underwater supersonic wave. The user interface of the network architecture is designed for collaborative use with researchers in other organizations. The flexible network architecture is based on the web direct-access model for the rotation mechanism. The user interface is available for monitoring and controlling via a web browser that is installed in any terminal PC (personal computer). Previously the combination of two applications was performed not by a web browser but by the exclusive interface program. So a connection model is proposed between two applications by indirect communication via the DCOM (Distributed Component Object Model) server and added in the web direct-access model. A prompt report system in the TS measurement system and a positioning and measurement system using an electric flatcar via a web browser are developed. By a secure network architecture, DCOM communications via both Intranet and LAN are successfully certificated.
The purpose of this paper is to propose a method to infer a truck driver's intent to change lanes before initiation of its maneuver. Observation on a truck driver's behavior in a real world and ordinary drivers' behavior at a driving simulator indicate common features that the drivers' glance at the side-view mirror and speed meter becomes frequently when they are going to change lanes in a near future. Based on the observation, a four-level method is developed for detecting driver's intent to change lanes by using the information of eye-movement. The method is applied to data collected in an experiment. The result shows that driver's intent is detected at a high level in average 80 percent lane changes when the maneuvers of changing lanes are initiated. Investigations also suggest that a risky lane change may occur if driver's intent is not detected at the high level. These findings are important to realize an adaptive support system in lane changes.
A linear system with a generalized frequency variable denoted by G(s) is a system which is given by replacing the variable ‘s’ in the original transfer function G0(s) with a rational function ‘φ(s)’, i.e., G(s) is defined by G0(φ(s)). A class of large-scale systems with decentralized information structures such as a homogeneous multi-agent systems, which has a common agent dynamics h(s)=1/φ(s), can be represented by this form. In this paper, we investigate fundamental properties of such a class of systems in terms of controllability, observability, and stability. Specifically, we first derive necessary and sufficient conditions that guarantee controllability and observability of the system G(s) based on those of subsystems G0(s) and h(s). Then we show that the Nyquist type stability criterion can be reduced to a linear matrix inequality (LMI) feasibility problem. Finally, we apply the results to stability analysis of large-scale systems in three different fields and confirm the effectiveness of the approach as a general framework which can unify variety of results for homogeneous multi-agent dynamical systems.
The standard computer-tomography-based method for measuring emphysema uses percentage of area of low attenuation which is called the pixel index (PI). However, the PI method is susceptible to the problem of averaging effect and this causes the discrepancy between what the PI method describes and what radiologists observe. Knowing that visual recognition of the different types of regional radiographic emphysematous tissues in a CT image can be fuzzy, this paper proposes a low-attenuation gap length matrix (LAGLM) based algorithm for classifying the regional radiographic lung tissues into four emphysema types distinguishing, in particular, radiographic patterns that imply obvious or subtle bullous emphysema from those that imply diffuse emphysema or minor destruction of airway walls. Neural network is used for discrimination. The proposed LAGLM method is inspired by, but different from, former texture-based methods like gray level run length matrix (GLRLM) and gray level gap length matrix (GLGLM). The proposed algorithm is successfully validated by classifying 105 lung regions that are randomly selected from 270 images. The lung regions are hand-annotated by radiologists beforehand. The average four-class classification accuracies in the form of the proposed algorithm/PI/GLRLM/GLGLM methods are: 89.00%/82.97%/52.90%/51.36%, respectively. The p-values from the correlation analyses between the classification results of 270 images and pulmonary function test results are generally less than 0.01. The classification results are useful for a followup study especially for monitoring morphological changes with progression of pulmonary disease.
This paper focuses on two control methods which are constructed based on neurocontroller (NC) and genetic algorithm (GA) for the Acrobot. A switching controller is first introduced where an NC optimized by GA is used for the swing-up stage and a linear quadratic regulator (LQR) is applied for stabilization. Next, we show that it is possible to handle both control stages of the Acrobot by using only the NC, called global NC, with a different evaluation function for GA. This controller seems to be the first smooth control strategy for the Acrobot. In order to analyze the characteristics and verify the effectiveness of the proposed control methods, numerical simulations are implemented using different timing constraints. A comparison with a classical controller is also provided. Simulation results show that the proposed controllers work effectively.