Search

This research develops a walking promotion device. The device is worn by user’s forearm and leads to increase arm swing. The device’s motion principle is based on parametric excitation. The Authors manufactured the prototype named “linear motion device”, and evaluated amplification of arm swing in previous paper. The device could increase the arm swing by 10°. In this paper, the authors experimented whether the device can increase the stride and walking speed. First, the participants wore the device on the right forearm. Afterwards, the authors measured at left thigh angle for stride and gait cycle. Comparing two states of disable with enable, the stride and walking speed were not increased. Second, they wore the device on the both forearms. As a result, the stride was increased 0.13 m, and the walking speed was increased 0.20 m/s. Therefore, the effectiveness of this device in walking promotion was confirmed.

This research proposes a walking promotion device using arm swing induced by parametric excitation. Generally, amplification of arm swing promotes a walking. The proposed device is worn on forearm. It has weight, linear rail and motor. Parametric excitation occurs moving the weight in the longitudinal direction of the forearm. The feasibility of the system is confirmed by the simulation by modeling the upper limbs and proposed device as serial links mechanism. Additionally, this simulation is conducted so that designing the mass of the device and movement of weight. Based on the result, the mass of the device is 0.43 kg, the movement of weight is 0.14 m, and the amplification rate is confirmed at 129.27%. In order to verify the simulation in the real system, 2-link pendulum was prepared. Comparing with two states of disabled and enabled, the shoulder joint angle increased 11.83° in average. In addition to this experiment, an evaluation experiment with human arm was performed. As the results, compared with simulation, the device can increase the arm swing at 10°.

We demonstrate short-range terahertz (THz) wireless transmission using a high-gain and low-profile dielectric cuboid antenna (DCA) in the 300 GHz band. The compact DCA will support high-speed mobile user equipment (UE) in the next generation. The simple rectangular structure enables easy installation and we configures the current design to a standard waveguide (WR-3.4). The DCA was developed with polytetrafluoroethylene and has dimensions of 1.36λ × 1.36λ × 1.79λ (1.36 mm × 1.36 mm × 1.79 mm at 300 GHz). Antenna patterns of the DCA at 300 GHz were evaluated using a near-field electro-optic sensing technique. The far-field pattern calculated from the measured near-field pattern showed good agreement with simulation results. The calculated antenna gain was 15.5 dBi. The wireless transmitter configured with the developed DCA has been demonstrated in the 300 GHz band. The resultant data rate with amplitude-shift keying of 17.5 Gbit/s was achieved at transmission distance of 79 mm.

This research proposes the walking promotion device using arm swing induced by parametric excitation. Generally, the walking is promoted by amplification of arm swing. The proposed device is wearable on wrist such a watch. It consists of weight, link and motor. Moving the weight in the longitudinal direction of the forearm, it occurs parametric excitation. In order to design the link length and weight, the arm and device are modeled and simulated with their parameters. In the result, the weight of the device is set to 0.10kg, the link length of it is set to 0.02 m, and the target amplification angle is decided 1°. To perform a preliminary evaluation experiment, 2-links pendulum was made with following the simulation conditions. Comparing with the states of disabled and enabled the device, the shoulder angle increased 2.23° in average. In addition, an evaluation experiment with human arm was conducted. As the results, the device can increase the arm swing.

Ultraviolet-B laser diode at 298 nm was realized by developing the technologies; growth method of lattice-relaxed Al 0.6 Ga 0.4 N layer on AlN/sapphire template and III-group composition-graded p-AlGaN cladding layer. The threshold current, current density and voltage were 5.0 A, 25 kA/cm2, and 34.4 V, respectively.

A pincer nail is a disease where a nail becomes deformed. The origin is unclear. Patients with a serious pincer nail cannot walk normally. Serious pincer nails are often treated surgically but tend to recur. Early diagnosis is essential because it allows for complete recovery using a special wire treatment. However, most patients do not seek diagnosis until the pincer nail is advanced. This paper proposes a system to assist doctors in early diagnosis and verifies that the system can distinguish between patients with a pincer nail and healthy persons. The system measures the gravity center fluctuation and uses a neural network to classify the data and defines the feature of a pincer nail patient. The results indicate that that the neural network can classify patients and healthy persons using the gait while walking.

The dexterity of fingers is known to indicate a human health condition; thus, a finger force distribution measurement device measuring finger dexterity, essentially, the force, posture, and movement speed of each finger when holding an object in a vertical position, is essential. In this paper, a finger force distribution measurement system was developed with a load cell built into the thumb sensing part, and with a force sensor calibration function solving the sensor responsiveness issue, toward improving the measurement accuracy. Experiments were performed to verify the effectiveness of the device in measuring the finger force distribution with high accuracy.

Since smoke does not have inherent color and shape features, conventional image analysis methods for object detection do not work well as they are expected. Therefore, feature extraction for smoke detection tends to be difficult. In addition, image information of smoke has a large variation due to its transparency and it is heavily affected from its background environment. To address these problems, this paper presents a block-based smoke detection method using a machine learning approach. In this method, a gray-scale block-based smoke region discrimination is presented based on the recurrent neural network with Long Short-Term Memory (LSTM). An image sequence is treated as a block-wise time series and the LSTM network discriminates blocks which include smoke region by the motion of smoke regions. The LSTM is expected to handle a large variation of smoke information caused from smoke's properties. In the presented method, image sequence is divided into fixed-size blocks which are treated as time series data and the LSTM is trained to discriminate whether each block contains smoke. The presented method is evaluated by several image sequences and training conditions and it is confirmed that LSTM based method can work well to improve smoke detection results.

In this paper, a method for the improvement of the relationship between calculation time and recognition accuracy in deep learning is proposed. A major problem with respect to deep learning is that a large calculation time is required for higher recognition accuracy. Because of this problem, the implementation of deep learning in hardware and its application to real problems are limited. In this study, layer-wise adaptive rate scaling (LARS) variables are adopted to evaluate the necessity of the learning of each layer. When the variable of a certain convolution layer exceeds the threshold value, the learning for that layer is considered unnecessary; thus, the layer is skipped. When a layer recognized as the layer that does not require learning, only the lower layers below than that layer are learned in the next epoch. By adaptively skipping the layer, the calculation time is reduced. Furthermore, the recognition accuracy is improved. Consequently, the proposed methods accelerate the calculation time in VGG-F to achieve the highest accuracy for the top1 and top5 test accuracy by a speed up factor of 2.14, and 2.25, respectively. Moreover, the respective top1 and top5 test accuracy was improved by 3.0%, and 2.8% which obtained as the final accuracy. In addition, the operation process was reduced by approximately 39.0% and required bandwidth was reduced by 38.9%, when compared with the case of conventional full layer learning.

This paper presents a proposal of a data-parallel stochastic gradient descent (SGD) using delayed weight update. A large-scale neural network appears to solve advanced problems, but its processing time increases concomitantly with the network scale. For conventional data parallelism, workers must wait for data communication to and from a server during weight updating. Using the proposed data-parallel method, the network weight has a delay. It is therefore stale. Nevertheless, it gives faster convergence time by hiding the latency of the weight communication for the server. The server concurrently carries out the weight communication and weight update while workers calculate their gradients. The experimentally obtained results demonstrate that, in the proposed data parallel method, the final accuracy converges within degradation of 1.5% compared with the conventional method in both VGG and ResNet At maximum, the convergence speedup factor theoretically reaches double that of conventional data parallelism.

This paper introduces a method to adaptively choose a learning rate (LR) with short-term pre-training (STPT). This is useful for quick model prototyping in data-parallel deep learning. For unknown models, it is necessary to tune numerous hyperparameters. The proposed method reduces computational time and increases efficiency in finding an appropriate LR; multiple LRs are evaluated by STPT in data-parallel deep learning. STPT means training only with the beginning iterations in an epoch. When eight LRs are evaluated using eight parallel workers, the proposed method can easily reduce the computational time by 87.5% in comparison with the conventional method. The accuracy is also improved by 4.8% in comparison with the conventional method with a reference LR of 0.1; thus, no deterioration in accuracy is observed. For an unknown model, this method shows a better training curve trend than other cases with fixed LRs.

In this paper, a finger force distribution measurement system for hand dexterity is developed, newly. This system consists of a measurement device, wireless communication units and a display unit. Measurement data are expressed on the display of the PC in real time and are stored. The other almost devices for finger force and hand force measure the maximum force. Our developed device measure the each finger force distribution, when a cylinder form thing such as a cup or a pet bottle is brought up. Using the measurement system, each finger force of some subjects that is healthy and young are measured to confirm the function of the device. The validity of the device is verified by the experiments. It is clear that influence of little finger is large and the position of thumb is important in grab motion.

The pincer nail, a disease with deformation of the foot nail, is so problematic that a patient is difficult to walk for the pain in the case of heavily deformed. The cause of the pincer nail is not clear. Some researchers consider that the cause is external factor. Our group also guesses that the gait motion is one of the cause. This paper mainly describes the measurement system for the gait motion we developed. And the measurement result of the gravity center fluctuation, fluctuation of the acceleration of the waist and the motion tracking by the developed measurement system are shown. A part of the data set that are obtained by the system is analyzed by the self-organized map and cluster analysis. As a result, the trend of the patient with pincer nail is found out.