Focusing on differences in changes in maximum sound pressures of bass components between rock music and EDM, we performed an experiment for their affective evaluation when listening them with their vibrations. The experimental results clarified the affective differences of the two types of music.

View full abstract

We present visual and auditory stimuli simultaneously to give a user pseudo-haptic feedback. We examine effects of sound volume change on perception of grasping a virtual object with a bare hand.

View full abstract

We present an auditory cue when a user most deforms a virtual object in a virtual space through a head-mounted display. We examine whether the addition of the auditory cue affects softness and operation of the virtual object.

View full abstract

We have been performing the satisfaction assessment of TV programs made by using TVML in order to examine the feasibility of substituting the real TV programs with CG. In this time, we have extended the genre of mimicking TV programs and conducted evaluation experiments. The result shows that the quality of CG program has simply raised the viewer satisfactory and the feasibility of substitution.

View full abstract

We propose a new web service, news with reading assistance for foreigners. As the reading assistance information, we use simplified Japanese, hiragana for Chinese characters, and a Japanese dictionary for children. We conducted a preliminary inquiry of its usability for education. We have developed a production support system of the news.

View full abstract

We aim to provide a real-time, high-quality and low network load video surveillance system. In this paper, for an early stage, we evaluate an accuracy of human detection in different resolutions of surveillance videos by using state-of-the-art human detection algorithms: YOLOv2 and OpenPose. Evaluation results conclude that adaptive control of video resolution may improve precision value or recall value.

View full abstract

In recent years, texting while walking has become a social problem since it causes a lot of traffic accidents. In the previous work, we have proposed a method to detect texting while walking by convolutional pose machines. However, only a pose of texting while walking has been recognized. In this paper, we propose to incorporate grasped object recognition.

View full abstract

In this paper, we evaluate recognition accuracy of human motion by using a state-of-the-art pose estimation technology called “OpenPose”. We construct 3D character models as a training data by using 3D modeling software and train the 3D data by using Support Vector Model (SVM) or Neural Network (NN). Evaluation results indicate that NN provides a higher recognition accuracy compared to SVM.

View full abstract

We propose a method for recognizing Japanese sign language from a video. We integrate responses from hand gesture and vowel estimation to achieve accurate Japanese sign language recognition. Experiments demonstrate the effectiveness of our method.

View full abstract

This paper proposes the quantitative evaluation method of aging impression by image analyses such as principal component analysis and texture analysis. This will help to indicate the effect of beauty products and to advise makeup methods suitable for that person.

View full abstract

Nose area doesn’t have clear contours compared to other facial parts. This paper proposes an automatic extraction method of nose shape from facial image considering the features of gray level distribution and shape peculiar to nose area. The drawing style of nose that reflects extracted feature points is also discussed.

View full abstract

As a method of personal authentication, biometric authentication has attracted attention in recent years. A 2D Face Recognition with 2D data shot by the camera, sensitive to changes in brightness and face orientation, there is also a problem in terms of security because it can match with the photograph. In this paper, we propose a 3D face recognition method using Point Pair Feature (PPF) which is used to recognize objects in the field of robot vision. By changing the face area using the distance between the eyes, it is not influenced by the shooting distance and the orientation of the face. Also, since 3D data is generated by using the Depth image acquired from the 3D sensor, it is not affected by the change in brightness. Therefore, it is possible to recognize the face robustly. From the experimental results, even if the distance to the sensor and the angle changed, it was confirmed that the person can be authenticated with an accuracy of 100%.

View full abstract

Automatic recognition and understanding for manga content has been studied for the purpose of new e-comic services. In the previous research, we proposed character face clustering method using feature vector output from hidden layers in CNN. In this paper, we investigate the influence by removal of background areas.

View full abstract

Crystalline selenium (c-Se), which shows a high spectral sensitivity in the visible light region, is suitable for a photoconversion layer in imaging devices. We optimized the structural design of c-Se-based photoconversion layers, which are overlaid onto the CMOS circuits, from the results of IV characteristics measured under different directional incident light.

View full abstract

We studied blue-sensitive organic photoconductive films for RGB-stacked image sensors. C60-doped DNTT derivative films sandwiched between ITO electrodes were fabricated. The EQE was enhanced up to 70% at bias voltage of 6 V.

View full abstract

We have quantitatively analyzed noise characteristics of a 4-chip HDR camera and showed the noise degradation between two differently exposed systems at the switching point. Also, we have proposed a new method to reduce this visual impact of the noise degradation and confirmed that the pixel-by-pixel switching method with creating an overlap period is effective.

View full abstract

In this study, we developed an on-chip dual color fluorescence measurement system using a CMOS image sensor to measure FRET signal in vitro. Three layers of optical filter are applied, which consists of an interference filter and an absorption filter to remove excitation light, and fluorescent color separation (blue and green) layer. As a result, we succeeded to measure FRET signal measurement in vitro using the developed device.

View full abstract

We have started medical application of 8K ultra-high definition (UHD; 7680 × 4320 pixels) imaging technology, which is a 16-fold higher resolution than the current 2K high-definition (HD; 1920 × 1080 pixels) technology, to a new rigid endoscopic system for advanced rigid endoscopic surgery.

View full abstract

We developed a VR system using hammock attached pressure sensors, and performed experiments to evaluate the system. More than half of users could feel thrill, but remaining users couldn’t feel it. In this study, we improved this system to feel more thrill and performed experiments to evaluate it.

View full abstract

We suggested image conversion method from 8K images to FHD on the display of mobile devices. We performed a psychological experiment to obtain the preference of proposed method comparing with simple down-converted images. The result indicates the preference was affected on the size of object in the images.

View full abstract

We have studied head movement and eye movement when changing viewing location and viewing distance. As a result, we show that a mean gazing point position depends on a presentation picture and a viewing location, and a viewing distance. Therefore we studied about the influence that difference of resolution and size brought at the mean gazing point position. It was shown that resolution and size affected the mean gazing point position from this result.

View full abstract

We examined a change of the mean gaze distribution, a change of the head angle of rotation, and a change of the convergence eye movement to objectively evaluate the difference of the effect brought by 2 channels stereo and 5.1 channels sound surround.

View full abstract

In order to predict human perception of direction of motion in depth (MID), we build a model based on Inter-ocular velocity difference (IOVD). We evaluated the model ability, predicting MID direction tuning reported based on psychophysical results in the literature.

View full abstract

Bit error rate (BER) performance has been used in digital systems such as communication and storage technologies. It is generally supposed that the noise in digital system is an additive white Gaussian noise (AWGN). In this paper, BER performance under simulation is compared with that under theory.

View full abstract

A training scheme for convolutional neural network (CNN), which can reduce bit error rate (bER) for holographic data storage, was studied. We evaluated the correlation between bERs with hard decision demodulation method and required training steps to shorten the training time of CNNs.

View full abstract

As a study of the presentation pattern used for pedestrian navigation, we examined the display style of LED indicator, color and figure pattern.

View full abstract

We present beam steering characteristics of an optical phased array (OPA) with electro-optic (EO) polymer waveguides. We obtained the beam deflection capability of the fabricated EO polymer OPA as 3.2 degree with applying voltage difference of 40V in adjacent phase shifters. Then we evaluated the half-wavelength voltage of the EO polymer phase shifter as 57V in the fabricated OPA.

View full abstract

For better motion image quality of OLED displays, we previously proposed an adaptive temporal aperture control method. In this report, we experimented for investigating lifetime degradation of OLED under the control, and discussed the results with a numerical analysis. The studies demonstrate the availability of our method.

View full abstract

We have developed a RGB Laser Backlight that uses three laser of red, green, and blue. The color gamut of the developed display nearly corresponds to that of BT.2020 using the laser for the light source.

View full abstract

We have developed white LED backlight that uses cylindrical lens. This type of backlight is easy to increase only LEDs without increasing the number of lenses, so it realizes high brightness that can correspond to 4K or 8K at low cost.

View full abstract

In November 2017, based on the fundamental technology of Digital Projection (Delta Group), for which they are known, Astro Design Co., Ltd. and Delta Electronics, Inc. will introduce the world's first DLP 8K projector. This report will introduce these outlines.

View full abstract

In this paper, we introduce a system to support tactical analysis by acquiring position information of players and a ball from monocular camera images and mapping them on the field for game video of rugby. In addition, automatic play classification using deep learning is also possible. We applied the proposed method to the actual game video and confirmed its effectiveness.

View full abstract

This research proposes a tool to calculate an optimum try route by solving an optimization problem based on reachable areas of players at each time. This tool also visualizes players’ movements with the optimum try route by using 3D CG from various viewpoints.

View full abstract

We propose a method to identify bike racers in race videos based on the posture information obtained by Realtime Multi-Person 2D Pose Estimation. We observed an accuracy improvement on bike racer identification by removing abnormal values of joint position information from SVM training data.

View full abstract