This paper proposes a method for determination of adjacent nodes to facilitate the formation of power efficient wireless sensor networks that are robust to channel variations caused by movement in the surrounding communication environment. Whereas conventional methods determine adjacent nodes in wireless sensor networks using one ID packet, our proposed method uses several ID packets that are collected in link investigation cycles. This method improves the battery life of the overall network system and thereby facilitates highly-reliable communication networks.
Various exogenous elements are involved in the thermal environment of the residential indoor spaces of building. Most of these elements behave differently and inter-relatedly to affect each room. It is difficult for the building air conditioning system to control the thermal environment of the rooms. Making use of the stability brought by the inherent equilibrium of the heat inertia system itself and regarding the changes by the result of various extrinsic factors as the change of the heat inertia of the room, the model of the differential air conditional system (DACS) is proposed. In order to verify the effectiveness of the system, the differential air-conditioning controller (DAC) is installed. It employs the rhyming control method. It is a method of air conditioning that causes gentle oscillatory movement to the process value of the room. In the controller, heuristic learning method is employed to determine the feed forward control parameters to capture the thermal behaviors (Differential Heat-effecting Dynamics), which is observed in the thermal environment of the room, defined as the thermal inertia. DAC keeps the appropriate temperature without heat shock to the residents in accordance with the trajectory of the temperature set point.
Remote surveillance of large-scale equipments such as power plants and building complex is important to prevent serious attacks and troubles. Automatic human action recognition can reduce the burdens of the surveillance. Multi-view video is useful for human action recognition, because it provides robustness to the changes of people's appearance by orientation and occlusion. One problem of conventional multi-view action recognition is that it requires both detection and tracking before action recognition. Human pose and motion vary depending on the person's action, and such variances may cause detection and tracking error. To solve this problem, previous work has proposed simultaneous action recognition and location estimation for single-view videos using Hough voting. In this paper, we extend the Hough voting approach to simultaneous multi-view action recognition and location estimation. Our proposed method independently casts votes for the action labels and spatio-temporal reference positions of actions in each view and integrates them using homographical transformations in the multi-view extension. We evaluated our method and confirmed that it achieved high accuracy in action recognition and location estimation. The contribution of this paper is that it enables multi-view action recognition without prior human detection and tracking.
Anomaly detection method from multi-dimensional time-series sensor data has been developed which detects anomalies based on normal state models. LSC method was employed to deal with various normal states and fast LSC method was proposed which reduces a computational time. Clustering is utilized to reduce the data for searching in FLSC method. Availability of FLSC method was confirmed using data of real equipment. FLSC method was 1 to 10 times faster than LSC method.
Eddy-current testing (ECT) technique, which is one of non-destructive evaluation, is applied to detect defect on printed circuit board (PCB). The ECT probe consists of a meander coil as an exciting coil and a giant magneto-resistance (GMR) sensor as a magnetic sensor. The ECT signal archived from a GMR sensor composes of many unwanted signals. Therefore, a defect detection system eliminates noise from measured ECT signal. If noise elimination requires many signal data, the probe necessary to move slowly in order to measure many signal data. Required signal data size depends on noise tolerance of defect detection. These mean that noise elimination is important factor in realizing high-speed defect detection system. We tried to eliminate noise in the time space, but it is difficult to remove the noise depending on the surface condition of PCB. In this report, we propose a defect detection method which eliminates noise not only in the time domain but also in the space domain. On a result, the method eliminates noise clearly by using smaller signal data and gives high-speed inspection of detect on PCB.
Imaging mass spectrometry using matrix-assisted laser desorption/ionization has increasingly got attention as a new analysis method of pharmacokinetics to accelerate the drug discovery processes. However, the detection sensitivity has been too low for practical analysis. By mixing zeolite (NaY5.6) with a conventional organic matrix, α-cyano-4-hydroxycinnamic acid (CHCA), the ion signal intensity of a drug, buspirone, at a concentration of 100 ng/mL was enhanced to about 1.6 times compared with that measured using CHCA only. The zeolite matrix enabled the detection of buspirone at a concentration of 0.1 ng/mL. The ion signal intensity of buspirone obtained from a brain section of a mouse administered buspirone was also enhanced to about two times by using the zeolite matrix compared with that measured using CHCA.
Femto-second pulse radiolysis system is developed for the study of the electron beam induced reaction in materials. Femtosecond electron pulse was generated by optimizing the laser photocathode RF electron gun accelerator. 240 fs time resolution was achieved by femtosecond electron pulse and formation of hydrated electron was observed by femtosecond pulse radiolysis. For more fine time resolution of the pulse radiolysis, the Equivalent velocity spectroscopy (EVS) method is required to avoid degradation of time resolution caused by velocity difference between electron and light in sample.
This paper proposes a novel method for designing linear phase FIR digital filters with asymmetric transition band. The proposed filters have a form of piecewise exponential functions in frequency domain where transition characteristics is determined by giving only two fixed feature points. The design method was evaluated in two ways. First, we evaluated a designed asymmetric transition band filter, and the results revealed that evaluated errors of the proposed filters are equivalent to a well-known existing filters. Then, it was shown that three-channel filterbank with an asymmetric transition band and liner phase property can be implemented.
This paper proposes a similarity measurement technique for retrieving similar driving scenes, using driving behavior signals and features of the driving environment. A previous work proposed a similarity-based retrieval system for finding driving data, which retrieved driving scenes by measuring similarity between scenes using driving behavior signals, such as steering angle and vehicle velocity. However, driving scenes can also be characterized by the surrounding driving environment. In this study, we assume that driving scenes consist of three major entities: the driver, the driver's vehicle, and the driving environment. We measure the distance between driving scenes using road features as well as the position and motion of surrounding vehicles (i.e., the surrounding driving environment), in addition to driving behavior signals obtained from the driver and the driver's vehicle. We then conduct a driving scene retrieval experiment to evaluate our similarity measurement method, using driving data collected on an expressway. Experimental results show that the additional use of environmental information significantly improves the precision of retrieval of scenes of driving events compared with a conventional method. According to our results, we also find that different people focus on different elements when comparing driving scenes, which may indicate that different drivers focus on different things when driving.
The method that binarizes an image by local thresholds in separated image areas is useful when the image has uneven brightness. And the method was proposed that separate an image dynamically for local thresholding to generate a binary image using a genetic algorithm. However, existing method used only horizontal and vertical lines for separating an image. This paper suggests a method to separate an image dynamically using diagonal lines for local thresholding to generate a binary image using a genetic algorithm and to evaluate binary image. The experimental results show that the binarized images by the proposed method is well separated object from the background, and that include less noise and blurs than existing method.
These days, there is more demand of camera based gaze estimation method for a new interface and a new marketing measurement tool. Considering these applications, the system should track a new user without any operation like calibrations. It also admits user's natural head pose changes. Previous methods, however, need calibration procedure before execution and have less accuracy under head moving situation. In this paper, we propose the method which tracks user's eyelid and iris automatically and accurately. Our method is the pretreatment of gaze estimation without any calibration and head pose restraint. First of all we track the facial feature points from an input face image and estimate its head pose, extracting eye region image. On the eye region image, we track eyelid shape based on the eyelid shape model generated beforehand from PCA. Finally we track iris inside the eyelid based on the eye ball model. These eyelid and iris tracking are processed by Particle Filter. From the evaluation of database including head pose changes, we confirmed that accuracy of the eyelid and iris tracking is improved compared with previous methods.
We propose an accurate ego-localization method by searching a streetview database composed of single front-view in-vehicle camera images. Previously, we proposed an image distance metric reflecting the positional relation of two cameras based on epipolar geometry analysis, and used it for ego-localization. However, since the method employed a dynamic time warping strategy to avoid the effect from outliers, both input and database images needed to be image sequences. To overcome this problem, the method proposed in this paper reformulates the previous image distance metric to a novel image distance that requires only a single in-vehicle camera image as an input, which is realized by considering the sequential property of the images in the database. We conducted experiments using multiple image sequences captured under various conditions by using an in-vehicle camera mounted on the windshield of a car. The experimental results showed that the proposed method could achieve an accuracy of 89%, and we confirmed its effectiveness.
We propose a new daily activity recognition method that can learn an activity classification model with small quantities of training data by sharing training data among different activity classes. Many existing activity recognition studies employ a supervised machine learning approach and thus require an end user's labeled training data, this approach places a large burden on the user. In this study, we assume that a user wears sensors (accelerometers) on several parts of the body such as the wrist, waist, and thigh, and we attempt to share sensor data obtained from only selected accelerometers (e.g., only waist and thigh sensors) among two different activity classes. For further reducing the burden on the user, we also adopt a semi-supervised approach.
This paper proposes the expat assessment system for global management personnel training utilizing the data mining employing evolutionary computation. This study follows the three steps. First step is to research those overseas temporary staff by direct hearing and having them answer the questionnaires. The second step is to make the data base with the results acquired from the step one. The third step is to construct the system to assess the prospective staff's aptitude or adaptability. The system would enable to promote the efficiency of assessment and to be useful in their training. This expat assessment system is built with the decision tree of evaluation standard based on the assorting problem solving method. For designing the non-terminal nodes of the decision tree, a new method is proposed utilizing the idea of the association rules, which is a powerful method in data mining. This is done to find out direct factors related to aptitude or suitability for overseas staff. The terminal nodes of the decision tree are to be optimized by Genetic Algorithm (GA) regarding a series of terminal nodes as the individuals forming the linear structure. As a result, the significant factors are found, which are strongly related to aptitude, adaptability and likeliness to be successful overseas. This research has revealed, by applying the system in practice, actual aptitude, skills and personalities required for expats.
This paper proposes a calculation method of local bifurcation points for discrete-time dynamical systems with piecewise nonlinear characteristic (PNDDS). First, an n-dimensional PNDDS, which has two piecewise nonlinear maps, is shown and its variational equation is derived. Next, the calculation method of the local bifurcation points which utilizes the conditional equation of the periodic solution and the characteristic equation is proposed. Then, we have to calculate derivatives of the map with an initial value and with a bifurcation parameter to obtain the bifurcation points continuously in the parameter space. The above calculation process is a key of the proposed method, and we explain it in detail. Finally, we apply the proposed method to a two-dimensional PNDDS and calculate the local bifurcation points for confirming its validity.
This paper proposes a new description method focusing on relations between businesses. This method is called Global Relations Diagram of function and demarcation (G-RD). In the computer application design, Current description methods utilizing ovals and allows like Data Flow Diagram (DFD) were widely used until now. However there are two problems with these description methods. Firstly, when this description method is utilized for a large-scale object, the arrow where it connects between the elements becomes entangled. Secondly, designed documents are not able to merge with other designed documents. G-RD can solve these two problems. It does this by describing the structure of the businesses by using “Elements” and “Relations”. Elements show the role and the function of the businesses, and Relations connect and explain the relationships between Elements. When Elements are plotted in the diagonal of square matrix, Relations are described at the intersections of the column and the row of two Elements. This paper introduces some case studies which were applied G-RD to design business processes, and to clarify the role and business assignment of each company after M&A. The effectiveness of G-RD has been evaluated at actual business field through over two hundreds of projects performed by Hitachi Ltd.
We have developed a new method of detecting utility poles in urban area by analyzing 3-dimensional point cloud acquired by the mobile mapping system (MMS). The MMS has laser sensors on its vehicle and surveys wide area using the sensors with driving in urban scenes. The output is point cloud, a large set of 3-dimensional points which represents 3-dimensional model of the area. Suppose a pole stands vertically in the scene, its measured points are located on its cylindrical surface. Therefore, if points gathered in a vertical cylindrical area, there must be a pole. In our method, such areas are searched by 3 steps, (1) projecting point cloud to a horizontal plane, (2) smoothing the projected pattern, and (3) matching it with the template which has a desired pattern of pole's projection. In this paper, we show our method and some experiments. Some extensions using ground filtering and the vehicle's locus are also introduced. In actual point cloud, poles are detected successively.