Deterministic oracle, used in standard software testing, is not available for statistical machine learning computer programs. Alternatively, metamorphic testing, pseudo oracle in view of data diversity, is adopted. Statistical machine learning, however, works on a dataset consisting of many data. Their distributions in the dataset introduce the diversity to testing. This paper proposes a new metamorphic testing method, in which generating follow-up test cases takes into account the dataset diversity. The method was applied to two well-known machine learning of support vector machines and neural networks.

View full abstract

In this paper, we propose an implementation method of hygienic macro-definition mechanisms into languages defined using OMeta, an OOP language with PEG-based general-purpose pattern matching. From the specification of a macro-definition mechanism and a parser written in OMeta, the proposed method generates an extended parser equipped with the macro-definition mechanism based on the Scope Set model. We show the effectiveness of the method by applying it to two different languages: JavaScript and MinCaml.

View full abstract

In this paper, we present a sphere-shaped interactive display system, named Qoom, as a new input and output device of the future. Unlike existing sphere-shaped displays, Qoom has perfect sphere shape, like a ball, which can be rotated, thrown or even kicked. At first, we discuss how sphere-shaped displays can be used in our life, and conduct preliminary user studies on how users interact with the sphere. Then we show how we developed a prototype of Qoom using touch and rotation detection, real-time object tracking and spherical projection mapping. We implemented primitive actions with the ball, such as touching, rotating, bouncing and throwing, as controlling actions in Qoom. We also developed applications for Qoom which utilized advantages of the spherical display.

View full abstract

We propose an immersive virtual reality environment configured in a swimming pool. This environment realizes various forms of underwater entertainments and enhanced swimming training environment. However, we have to overcome some technical challenges caused by water and surround-screen. In this research, we focused on investigations on issues related to infrared (IR) radiation absorption and reflection on the walls of the swimming pool and proposed the solutions. The difficulty on position tracking is that existing motion capture systems using IR or blue lights are not feasible for an underwater immersive projection environment. The reason is IR is absorbed, and visible lights are unstable in complex backgrounds. Therefore, we propose a circular-polarization-based technique for position tracking using visible-light LED markers. In addition, we also took investigations on an optical distortion in the water, and reflections causing a decline of immersion.

View full abstract

Cardboard virtual reality (VR) environment using a cardboard viewer and smartphone as a head-mounted display (HMD) opens the door to new possibilities where people can experience VR easily and inexpensively. However, such cardboard VR environment covers the touch screen of the smartphone and thus restricts the user from touching the screen. To address this problem, we present FistPointer, a target selection technique using mid-air interaction behind a smartphone for cardboard VR environment. Our technique displays a pointer on the screen corresponding to the position of the hand detected by the built-in back camera of the smartphone. The user can move the pointer by moving the hand in a thumbs-up posture and select a target by folding the thumb similar to pushing the button of a joystick. Our technique can be implemented using only the built-in camera of a smartphone, thus it is easy to apply our technique to a target selection in cardboard VR environment. To evaluate the performance of our technique, we conducted experiments in selection task. Furthermore, we developed a game using our technique and investigated the user impressions.

View full abstract

In this study, we propose a support system for solo live reporting. Personal live reporting has become popular because of widespread use of live streaming services such as Youtube Live. In solo live reporting, a reporter will be struggling to deal with a lot of work, particularly because he/she cannot edit the report after shooting. To overcome these difficulty, we designed and implemented the support system for solo live reporting. We also evaluated the system through the both qualitative/quantitative aspects and revealed that our system can deliver live reporting as same quality as the traditional form of live reporting (i.e., live reporting by the team).

View full abstract

In this paper, we propose a new apporach to proving loop termination that exploits temporal information such as “the value of this expression will always become positive (negative) eventually during iterations”. Considering higher order differentials, it is possible to handle the changes in the values of expressions over multiple consecutive iterations straightforwardly. The proof can be formalized in the ordinal first-order logic framework, and can be automated efficiently using SMT solver through the reduction to satisifiability problem.

View full abstract