We have examined the non-contact respiration measurement by applying the active stereo for realizing easy screening of Chronic Obstructive Pulmonary Disease (COPD). In this study, we apply the Self-Organizing Map (SOM) for classification of unforced respiratory waveform. The automatic classification of the respiratory waveform is based on the difference in the undulating pattern of respiratory waveform between the COPD patients and the healthy subjects. As the result of the experiment, it becomes clear that the classification by SOM is almost equivalent to the classification by respiratory specialists.

抄録全体を表示

In this paper, we propose a gesture recognition system to operate a mobile robot as one of a natural user interface between man and machine. In this system, five gestures are used for the operation: one of the gestures is to stop robot, and the others are to move it. These gestures are shown by the continuous movements of the human right hand. The Kinect sensor is used to capture 3D coordinates of right hand. And, we use the DC components and the differences of phases as features to recognize gestures. These features are estimated by the Fourier transform of 3D coordinates. The input gesture is recognized by the similarity using these features and the learning data. In the experimental results, when the input gesture changes to the other one, the proposed system often recognizes the previous gesture. But, we can show possibility that our system can recognize gestures to operate the mobile robot.

抄録全体を表示

This paper introduces a novel method to automatically remove electro-oculogram (EOG) from electroencephalogram (EEG) using second-order blind identification (SOBI) and nonnegative matrix factorization (NMF). To demonstrate the effectiveness of the proposed method, we applied SOBI, NMF, or combined SOBI and NMF (SOBINMF) to the EEG data during motor imagery task in which participants imagined moving their left or right hand. Ocular artifacts were removed more effectively in our proposed SOBINMF method compared to SOBI or NMF alone.

抄録全体を表示

Insulin secretion from pancreatic β cells plays the central role in maintaining glucose homeostasis. Impaired insulin secretion mainly contributes to the pathogenesis and pathophysiology of type 2 diabetes mellitus. Diabetes of the type-2, which are typical among Asian patients, are caused by insufficient insulin secretions by unknown factors. For revealing the factors and dealing with them properly, it is necessary to clarify the insulin secretion mechanism in pancreatic β cells. In insulin granule kinetics, i.e. the transportation process of insulin granules produced in pancreatic β cells to the cell membrane for secretion, the granules are supposed to move along the cytoskeleton. For constructing the simulation model, we divided the model into four stages based on the skeleton of a pancreatic β cell and the principle of granule kinetics, i.e., the granule producing site, the inner layer, the outer layer and the granule secretory part. The phenomena to be reproduced are the typical secretion patterns found in healthy person, diabetic patients and healthy person with incretins. In this paper, a series of simulations based on the data concerning insulin granules kinetics have been carried out, where the typical secretion patterns have been reproduced by setting the parameters adequately according to the observation in a pancreatic β cell without changing the model itself. Through these simulation results, the potential of the proposed model has been confirmed.  

抄録全体を表示

A small subset of neurons, called “leader neurons,” has been assumed as the sources of network bursts in dissociated neuronal cultures. In this paper, we proposed a network burst generation model that a network burst is considered as a sequential transition of spatial activity patterns lead by a “leader pattern”. We recorded spontaneous activities of cultured cortical networks with high-density CMOS microelectrode arrays. Spatial patterns were extracted from the high dimensional recorded data using non-negative matrix factorization (NMF). Then, we hypothesized the state-space model where the leader pattern served as input and the others served as states, respectively. After estimating the model parameters from the training data, we attempted to restore the activities of test data with the estimated model. As a result, the spatio-temporal patterns in network bursts were successfully reconstructed from the model, suggesting that the leader pattern is a crucial predictor of the network burst.

抄録全体を表示

In this paper, we propose a new H_∞ controller reduction method using the block balanced truncation with the Riccati equations. To maintain the closed-loop properties such as stability and H_∞ performance index, we employ the Riccati equations for controller reduction. At first, an H_∞ controller is designed using a generalized plant with some frequency-weightings. Then we construct a new virtual generalized plant which consists of the controller and the frequency-weightings. A low-order controller is obtained by balancing the solutions of the Riccati equations which are derived from the virtual generalized plant. A numerical example is illustrated to show the effectiveness of our method.

抄録全体を表示

An active noise control system using the gradient based adaptive algorithm without a model of the secondary path is proposed. The cost function of this adaptive system is a square sum of the cross-correlation coefficient of the input noise and the output error signal. Importantly, this system is composed of single transversal adaptive filter that estimates the inverse transfer function of the secondary path and the transfer function of the primary path at the same time. A new gradient based adaptive algorithm and its fast version using only the input signal and the output error signal are formulated, then the above mentioned analytical results are verified by computer simulation.

抄録全体を表示

Drivetrain bench system is utilized to test powertrain components of a vehicle such as transmission, torque converter, etc. Conventional drivetrain bench system emulates the moment of inertia of a vehicle body. In order to simulate real road loads of the powertrain components, it is necessary that the dynamometer of the drivetrain bench emulates the moment of inertia of a tire. The moment of inertia of the tire is lower than the moment of inertia of the dynamometer. So, the dynamometer needs low inertia emulation control. The drivetrain bench is mechanical vibration system. The principal factor of the vibraion is stiffness of a drive shaft. In this paper, we propose a low inertia emulation control method for vibration system and its application to simulating real road loads. Basic idea of the proposed low inertia emulation control method is feedforward of torque that is necessary for accelerating emulated inertia. By numerical simulations, it is shown that real road loads tests can be implemented with the proposed control method.

抄録全体を表示

Human segmentation is one of the most interesting yet most challenging subjects in the field of object segmentation and image processing. It can be used in various types of applications from image retrieval to robotics and human machine interfaces, including even entertainment. Many researches have been done on this subject and it is still one of active research areas. But until now, a method for accurate segmentation in different conditions has not been introduced. In this paper, we present “Statistical Shape Feedback Segmentation” (SSFSeg) method, which is a way to automatically segment human subjects (pedestrians) from single images. Our main contributions in this paper are: 1) Using human shape model as priors for Grab-cut segmentation. 2) Implementation of a feedback system which provides a coarse-to-fine way of generating more accurate shapes. For this task, we try to use masks generated by the Statistical Shape Model (SSM) algorithm as a prior input for the Grab-cut technique to segment the desired human subject in the image without user interaction. To achieve this, we propose a feedback framework for the SSM sample generation. Our experiments confirmed that the segmentation error of our proposed method is less than half of the Grab-cut method.

抄録全体を表示

In this paper, we propose a speaker indexing method using speaker verification technique to extract one desired speaker's utterances from conversational speech. To solve the overlapped speech problem, we construct overlapped speech models with the observed conversational speech itself. The overlapped speech models include overlapped speech of target and cohort speaker, and speech model of two cohort speakers. In order to evaluate the proposed method, we made a simulated conversational speech that has up to 50% overlapping segments. The EER was reduced by up to 43.7% compared with the conventional methods that use a target speaker model only, and use a target model and overlapped speech model trained with a speaker independent large speech database.

抄録全体を表示

The opportunity to use smart devices in business operations is increasing, because of the evolution of smartphones, tablets, high-speed Internet, and the spread of cloud services. In the development of Web applications utilizing smart devices, the productivity and the maintainability are very important. Automatic generation tools of Web applications are effective to improve productivity, but they are required to change and add business logics to auto-generated programs. And there are variety of OS, screen size, and operability in smart devices. Therefore it is necessary to develop the program correspondingly to each smart device. In this paper, we propose a method of automatic generation of Web application considering the variety of smart device. And we implemented a prototype of maintenance engineer's system to evaluate the proposed method.

抄録全体を表示

This paper presents a compact feed matrix circuit for use in multi-beam antennas which controls beam directions including boresight. The feed circuit consists of two rat-race circuits and a three-way switch, and it incorporates one input port and four output ports. Its switch setting allows this system to obtain phase differences between antenna elements (namely 0°, 90°, or -90°) and accordingly switch the beam in three directions. The feed circuit was designed at 2.45GHz using a microstrip line, and the characteristics of the fabricated circuit were evaluated. The measured results agreed well with those of the simulation.

抄録全体を表示

This paper proposes a compact rat-race circuit using LC lumped parameter elements. The circuit, which converted transmission lines into pi-type phase shifters, consists of six lumped parameter elements. Therefore, the circuit can be the size of 1/267, while having a conventional performance. The prototype for use in 1GHz band is fabricated on a substrate, and the propagation characteristics and reflection characteristics of each port are measured. The measurement results agreed well with those of the simulation. The dimensions of the circuit are approximately 5.5mm×5.5mm.

抄録全体を表示

This paper proposes the Copy-Learning-Based RBF network which is the combination of RBF network and a part of network and describes the effectiveness of the proposed network with a character recognition experiment. The experimental result has shown that the learning of the proposed network was more stable and its convergence was also earlier than conventional Copy-Learning-Based BP network, in addition the proposed network is superior in ability for generalization.

抄録全体を表示

This paper proposes surrounding lighting system to increase realistic sensation. Several tags are annotated to timeline in a video data and are used for controlling home appliances. In order to deal with configuration and installing situation of each user, we develop a method to relate annotated tags to user's environment. This paper illustrates the details of the method to harmonize several lighting and video by using ECHONET Lite protocol.

抄録全体を表示