The research on face image generation has gained widespread attention in the field of image generation. However, large datasets are required to train a generative model to produce images of high quality and resolution. It is especially difficult to collect multiple face images of a specific group. It may also be difficult to establish an efficient model by training with a small dataset.This paper presents a novel method for the generation of face images with the features of a specific group. The face images of a dataset are embedded into a latent space as sets of latent variables by Image2StyleGAN and are expressed as a distribution of the sets of latent variables in the latent space of a pre-trained StyleGAN2. Principal Component Analysis (PCA) is used to extract the features of the distribution. The generation of images for different groups does not require re-training of the model, as a pre-trained model is used instead. Furthermore, the proposed method can generate high-quality images with the features of the group from a dataset with only about 100 face images. However, the quality and the variety of the generated images can vary depending on a Cumulative Contribution Rate (CCR) of the PCA. Therefore, this study also proposes a metric called the Fréchet Inception Distance in Principal Component Space (FID-PCS), which can evaluate the generated images even with a small dataset. The FID-PCS can be used to determine the CCR that generates images with a good balance between the quality and the variety. The face images of three groups were collected as datasets to evaluate the validity of the proposed method, which include male idols, female idols, and male mixed martial artists. It was observed that images with the features of the group are generated by the face distributions extracted by the PCA, and the images with high quality and wide variety are generated by determining the appropriate CCR by the FID-PCS.

This paper introduces a display technique to show gradation animation using a shape-changing artificial grass system. Recently, liquid crystal displays (LCDs) and projectors have been used in a natural public facility such as a park. However, these displays sometimes spoil natural scenery because they look like artificial materials. Our proposed method is useful to develop a natural landscape-friendly display. We develop a grass pixel system to display the gradation. The grass pixel consists of yellow and green grass, and the length of the green grass moves vertically. The grass pixel uses a pin-based shape-changing system using a stepper motor to control the grass length dynamically. We evaluated the gradation of the grass pixel through an image processing of HSV (Hue, Saturation, Value of Brightness) values. In the measurement, the pixel system moved the grass length from minimum to maximum, and we took images of the running grass pixel from several angles. As a result, we observed that the HSV values changed linearly with the movement of the length of the green grass. We also developed a 3 × 3 pixels grass display, which can show several gradation animations as demonstrations.

This paper presents a novel workspace for sharing a virtual space among multiple users with tangible interfaces and multiple perspectives. The multiple users can share a virtual space such as a spatial design through first-person perspectives (FPPs) and third-person perspectives (TPPs), and can discuss it while switching between the FPPs and TPPs. The workspace, named the Intermediation Family, is composed of box-shaped movable tangible interfaces, a fixed large display, a simple removable HMD and a computer that makes a virtual space. Since the tangible interface plays as both a user's avatar and a camera object for the FPP in the virtual space, the user can observe the virtual space from the FPP on a display mounted on the tangible interface while moving it. When the user wants a more immersive experience, she/he can also use the HMD. Moreover, the users can observe the entire virtual space through the TPP on the fixed large display. The multiple users can share the spatial design by switching their perspectives and exchanging the tangible interface and the HMD one after another. In this paper, two applications were implemented to hear opinions of some users. One is a virtual walk-through application, the other is a hands-on room layout application.

In this study, we present a method for controlling the surface shape of transparent liquids that generate caustics due to luminance distribution of an arbitrary image. Our method consists of the process of caustics design and the process of liquid surface control. While designing caustics, an arbitrary grayscale image is treated as the input image and the surface shape of transparent liquid that generates caustics is calculated based on the luminance distribution of that input image. We utilize the Poisson's equation to obtain a continuous liquid surface. A driving force based on the current height field and the target height field is introduced as the external force. This force that configures the current liquid surface into the target shape in the previous process is calculated and the change in the water surface and the caustics due to this force is verified using fluid simulation.

When pedestrians and a mobile robot pass each other in a corridor, it is important for a robot to make an appropriate path in order not to confuse pedestrians. This paper presents a new path planning method where the mobile robot follows the rule of “eeping to the right.” Three kinds of maps based on laser range finder data are used in the path planning. One is a global cost map made by a simultaneous localization and mapping (SLAM) technique, that expresses a whole environment and is used to make a path from a start position to a goal. Another is a local cost map, that is used for avoiding obstacles near the mobile robot. The other is an attractive force cost map proposed in this paper, that makes the robot keep to the right along a corridor. When there is no pedestrian, the mobile robot goes to a goal position indicated by an operator based on the global cost map while avoiding moving obstacles based on the local cost map in real time. When the robot detects pedestrians, it begins to keep to the right. This method where the robot follows a social norm depending on each country can make a pedestrian pass by the mobile robot more smoothly. Some experimental results reveal the validity of our proposed path planning method.

This paper presents a new robot system consisting of plural small size rovers for an asteroid exploration. Each rover can communicate with others using radio, and a wireless mesh network is configured on an asteroid's surface. Our proposed system has the following three advantages against a conventional exploration system using one or two rovers: (1) It is possible to explore a wider area of an asteroid. (2) Since the mesh network has redundant communication paths, it has more robustness against some troubles. (3) It is possible to estimate the relative distances among plural rovers by using the mesh network. This relative distance estimation is useful for asteroid analyses using sensors that rovers have. Simulation results reveal the validity and effectiveness of our proposed rover system with the wireless mesh network and the relative distance estimation method.

This paper presents a guidance method using a laser pointer attached to a librarian robot and gestures consist of body movements. Our librarian robot, that is fixed in an experimental environment, is able to say greeting properly based on estimation results of human behavior using a laser range finder, talk with a library user with a natural language, search books depending on a user's request. When the robot explains where a requested book is in the library, it points out a target position using a laser spot while the body and head are also turned around to the direction of the target. This makes detailed oral explanations unnecessary. However, the robot cannot always point out a target position directly using the laser pointer. So we design three types of guidance using the laser pointer and the gestures depending on the library environment. Some experimental results reveal the validity and effectiveness of our proposed guidance method using the laser pointer and some gestures for our librarian robot.

Abstract This paper presents a new voice morphing method which focuses on the continuity of phonological identity overall inter- and extra-polated regions. Main features of the method are 1) to separate the characteristic of vocal tract area reso-nances from that of vocal cord waves by using AR-HMM analysis of speech, 2) interpolation in a log vocal tract area function domain and 3) independent morphing for the vocal tract resonances and vocal cord wave characteristics. By the morphing system constructed on a statistical conversion me-thod, the continuity of formants and perceptual difference between a conventional method and the proposed method are confirmed. Index Terms : morphing voice, phonological continuity, speech synthesis, inter- and extra-polation 1. Introduction Voice morphing is a technique for continuously modifying a source speaker’s speech to a target speaker’s, whereas voice conversion usually means transformation from a source speak-er’s speech to a target speaker’s. Therefore, in voice morphing, the interpolated voice is required to maintain phonological identity even between the two speakers’ utterances. The pre-sent research focuses on this aspect of phonological continuity in the overall interpolated section and also that of an extrapo-lated region, since that will be useful for such applications as creation of peculiar voices in animation films. Since 1990s, many techniques for voice conversion are proposed[1-7]. One successful technique is to use a statistical method for mapping from a source speaker’s voice to a target speaker’s in the cepstrum domain[2,3]. However, weakness of this type of methods is discontinuity of formants because the relationship between formant transitions and the time pattern of the power spectral envelope sequence is nonlinear, that is, continuous interpolation of log power spectra (i.e. cepstra) does not result in continuous formant transitions. For example, when one power spectrum has a peak (formant)

This paper reports a new myoelectric interface for robotic hand control consisting of two main parts. The first part concerns the motion classification using electromyogram (EMG) signals of a support vector machine (SVM). Because there has been little research on the application of the SVM to motion classification using EMG signals, its effectiveness has not yet been established. The SVM has some advantages with respect to generalization and computational complexity, and therefore, we used the SVM to examine its classification ability. The second part concerns the estimation of an operator's joint angle corresponding to the motion determined by the first part. Estimation of the operator's joint angles is based on EMG-Joint angle models, which express the linear relationships between the EMG signals and joint angles. To verify the effectiveness of our interface, we performed off-line hand motion classification and real-time robotic hand control experiments with eight subjects. The experimental results showed that seven hand motions achieved a classification rate of more than 90% for all subjects. In addition, a three-dimensional computer graphics robotic hand was controlled in real-time (62.5 Hz) without delay.

This paper describes a new observation system for monitoring the operation of a micro machine, and a controller based on the visual servoing method for controlling a micro machine. The application considered is the splicing of optical fibre ribbons. The proposed observation system has only one CCD camera. However, it can be converted into a stereo vision system by using two mirrors. By using the new controller, systems can be controlled based on visual information on image planes rather than in three dimensional space. Our two proposed methods result in a system that is highly accurate, simple and inexpensive. Experimental results on a prototype system reveal the validity and effectiveness of the observation system and the controller based on the visual servoing method.

Although visual servoing is suitable for control of a mobile robot based on a single on-board camera image, wheeled vehicles have a non-holonomic property that the degree of freedom of the input is less than that of the configuration. Thus the visual servoing can not be directly applied to a camera system fixed on the vehicle. Furthermore there is no smooth and stable state feedback law for non-holonomic vehicles. To overcome these difficulties, the authors consider another situation, that the pan angle of camera is actuated and used as an input of visual servoing. The desired state is given as not a constant value but a trajectory represented in terms of image sequences of a gazed object. Based on this idea, the authors propose a method of realizing "teaching and playback" for a non-holonomic mobile robot. In the teaching mode, the robot tracks a landmark in the environment by a single on-board camera and stores its image sequences while running commanded by a human operator. In the playback mode, the robot tracks the desired trajectory with velocity commands for the vehicle and the pan angle for the camera determined from the stored and current images. The authors discuss the convergence to a desired state and moreover point out an additional feedback term, of a deviation of the pan angle works to improve the tracking performance. The effectiveness of the proposed method is shown by illustrating computer simulation results.>