Search

In this paper, we propose a methodology of forecasting the direction and extent of volatility in mid-to long-term excess returns of stock prices by applying natural language processing and neural networks in the context of analyst reports. Analyst reports are prepared by analysts in the research departments of stock brokerage firms. We examine the contents of reports for useful information on forecasting the movements of stock prices. First, our method extracts opinion sentences from the reports while the remaining parts are classified as non-opinion sentences. Second, our method predicts stock price movements by inputting the opinion and non-opinion sentences into separate neural networks.

Image-based virtual try-on is an area of research that is attracting attention as the demand for online apparel shopping continues to increase. The methods proposed thus far have focused on how to generate a dress-up image while preserving the clothing details. However, the posture of the model in the image is limited to an upright position, and other positions frequently do not work well. In this study, based on a kind of generative adversarial network (GAN) that utilizes UV mapping to consider the 3D structure of the human body, we propose a novel virtual try-on method called a UV Try-On Network (UVTON). We use a DensePose to estimate a point corresponding to the 3D surface of a human model for each pixel point of a 2D image and incorporate the estimated information into our model. It is thus possible to change the clothes of users holding various postures. Our proposed method uses UV mapping and two other modules. One module generates parts to be used in the mapping, and the other refines the image and produces a more realistic image. Based on both qualitative and quantitative comparison with existing methods, we experimentally demonstrated that our method achieved better results with various postures.

It has been discussed that the increase in the intake ratio of calcium and magnesium (Ca/Mg ratio) increases the risk of the ischemic heart disease. Though there are some reports on Ca/Mg ratio concerning the ischemic heart disease in animal experiments, there are few reports of experiments on cultured cells. Therefore, we investigated the influence of various ratios of Ca/Mg on the cell viability of cultured fibroblasts at various passage numbers. As a result, cell viability of cultured fibroblasts at passage number 14 or more significantly decreased with rises in Ca/Mg ratios. Furthermore, it was shown that this phenomenon was related to Ca/Mg ratios and not to rising Ca concentrations used in this report. Additionally, it was shown that the supplementation of DSW collected from Izu-Akazawa, Japan which did not change the Ca/Mg ratio, to cells in a state of decreased cell viability due to the increased Ca/Mg ratios showed an inhibition against decreasing their cell viability. Considering these results, it is suggested that the application of DSW will be a promising approach on health maintenance because the Ca/Mg ratio has been an increasing trend along with the diversification of the Japanese diet.

We evaluated the exactness of interactions using our proposed technique for applications in which users' bodies interacted with the virtual objects presented with 3-D displays in front of them. Interactive applications require interactions to be executed when users' bodies are at the positions of virtual objects; however, conventional techniques that execute interactions when the bodies are at the positions calculated from the binocular disparity make it difficult to meet this requirement because the calculated positions are often different from the positions of virtual objects. The proposed technique estimates the positions of virtual objects by using human body movements, and meets the requirement by executing interactions when users' bodies are at the estimated positions. We conducted an experiment in which users actually interacted with virtual objects, and the results indicated that more interactions succeeded when the proposed technique was used than when conventional techniques were used. We demonstrated that interactions using the proposed technique were exact.

Ternary content addressable memory (TCAM) is popular LSI for use in high-throughput forwarding engines on routers. However, the unique structure applied in TCAM consume huge amounts of power, therefore it restricts the applicability to deployment for handling large lookup-table capacity in IP routers. In this paper, we propose a commodity-memory based hardware architecture for the forwarding information base (FIB) application that solves the substantial problems of power and density. The proposed architecture is examined by fabricated test chip with 40nm embedded DRAM (eDRAM) technology, and the effect of power reduction verified is 95% lower than conventional TCAM based.

A novel 100G high gain FEC performance measurement scheme employing simple frame multiplexing and synchronous data mapping of 10Gbps PRBS signal is proposed. We successfully demonstrate FEC performance measurements around 2×10−2 input BER condition.

We experimentally investigated the combined effect of DGD and PDL in installed fibers for the first time. The experimental results show that the Q-factor variation caused by PDL decreases as the DGD value increases.

This paper proposes a fast, precise transient response and frequency characteristics simulation method for switching converters. This method uses a behavioral simulation tool (MATLAB/Simulink) without using a SPICE-like analog simulator. The nonlinear operation of the circuit is considered, and the nonlinear function is realized by defining the formula based on the circuit operation and by applying feedback. The transient response and frequency characteristics of a current-mode boost DC-DC converter and a charge pump that were designed using a 0.18-um CMOS process were simulated by SPICE and the proposed program on a behavioral simulation tool, which we named NSTVR (New Simulation Tool for Voltage Regulators). Simulation results showed good agreement, yet NSTVR was more than 85 times faster than SPICE in CPU time in calculating frequency characteristics of a current-mode boost DC-DC converter.

In order to treat the virtual object in the same way as the real object, the authors propose the new technique for the prediction of the visually perceived location. In mixed/augmented reality, the real and virtual objects coexist. Because the real object is directly treated with the observer's body, if the virtual object is not, the observer is confused. Although the best way for direct treatment of the virtual object is that a system of mixed/augmented reality detects the visually perceived location, it is impossible. In the authors' proposal, the visually perceived location is predicted using the observer's action. Such proposal is evaluated by examination of the prediction of visually perceived depth using action that the observer reaches out for the virtual object. The results suggest that the visually perceived depth is predicted by fitting the depth of the hand as a function of action time into the logistic function.

Described are studies on human factors of 3D image displays for virtual reality. Firstly, research into human factors of the immersive multi-projection image displays is given to characterize the visually induced self motion perceived by human eyes. Secondly, a multiple-angled 3D video display system is given, by which an audience is permitted to enjoy musical plays performed in a concert hall at a distant location in real time. Thirdly, a new 3D image display is given, on which differences between the depths of objects can be perceived due to motion parallax, a cue for sensing the depths.

We propose a novel technology that can prevent the illegal use of images of objects that do not have watermarking. This technique uses illumination that invisibly contains the watermarking. As the illumination for the object contains watermarking, the image of the object taken by the camera also contains watermarking and this watermarking can be extracted by image processing. We conducted experiments where one-bit binary data were embedded in one block that consisted of 8 times 8 pixels using the phase of the highest frequency component generated by DCT. The experimental results revealed that embedded data could be read out extremely accurately, reaching 100% with the majority method, which used three data embedded in three separated blocks. The experimental results demonstrated the feasibility of the technology we propose.

This paper describes a 3-D display that can express differences between depths at extended distances of over a hundred meters for moving observers using motion parallax. The results of a subjective test demonstrated that the perceived depth of the image at long distances could be controlled using artificial motion parallax generated by changing the rate of expansion of the image size, irrespective of its real depth. As a result, we demonstrated its possibility for applications to car-navigation systems combined with augmented-reality technology.

This paper proposes a new 3D display that can express differences between depths at extended distances of over tens of meters to meet new requirements for outdoor use. We attempted to use motion parallax for observers to perceive depth because this works as a cue at longer distances where binocular parallax, which is used in conventional 3D displays to perceive depth, does not work. We conducted subjective tests using a moving car in which observers viewed a test pattern overlapping the real view ahead of the car seen through the windshield to examine the feasibility of the 3D display we propose. The experimental results revealed that the perceived depth of a pattern could be controlled by changing its rate of expansion, demonstrating the feasibility of a 3D display for extended-depth perception using motion parallax.