Gaze input for electric wheelchairs offers an alternative to joysticks, but conventional dwell-time methods suffer from user fatigue and the "Midas Touch" problem. Furthermore, their reliance on infrared (IR) cameras leads to high costs and poor robustness outdoors due to sunlight. This study proposes an IR-free system combining the Eye-Glance gesture method with a low-cost, visible-light (RGB) camera. While our prior work introduced this concept, its simple algorithm was limited to indoor use. We overcome this by introducing a new pupil-center detection algorithm based on image saturation, shape constraints, and blink handling, specifically designed to tolerate ambient light changes. Evaluation with 8 participants performing four gestures under indoor and outdoor conditions achieved average first-attempt success rates of 95.9% (indoors) and 94.4% (outdoors). This work demonstrates that our visible-light pipeline provides a low-cost, robust gaze interface suitable for electric wheelchairs in varied environments.

In this paper, we propose a simple method for annotating reading order based on the measurement of manga reading behavior, which automatically estimates reading order by obtaining a scanpath from eye movements measured by an eye tracker with a built-in HMD. Our method reduces the cost of work and measures natural reading behavior, which is difficult to achieve with conventional GUI-based manual labeling and image analysis-based methods for manga reading order annotation tasks. First, Gaze samples captured with an HMD built-in eye tracker are mapped to their corresponding panels, after which time-series clustering is applied. Then, we extract the most extended fixation duration for every panel and sort panels in ascending order of that duration to infer an individual reading sequence. Also, as needed, it integrates by voting across multiple estimation results, then yields a consensus—i.e., tendentious— reading order for the page. In addition, we evaluated it on 99 pages selected from 33 commercially published works, using recordings from five adult readers. The automatically inferred sequences were compared against expert manual annotations, achieving a mean Kendall’s rank correlation coefficient of 0.82. This result indicates that the proposed method can be extended to annotation tasks based on more natural reading behavior while maintaining estimation accuracy. At the same time, the observed variability among individual sequences highlights the importance of modeling reader-specific behaviors in future work.

In this study, we discuss issues in the traditional evaluation norms of trend forecasts, outline a suitable evaluation method, propose an evaluation dataset construction procedure, and publish Trend Dataset: the dataset we have created. As trend predictions often yield economic benefits, trend forecasting studies have been widely conducted. However, a consistent and systematic evaluation protocol has yet to be adopted. We consider that the desired evaluation method would address the performance of predicting which entity will trend, when a trend occurs, and how much it will trend based on a reliable indicator of the general public's recognition as a gold standard. Accordingly, we propose a dataset construction method that includes annotations for trending status (trending or non-trending), degree of trending (how well it is recognized), and the trend period corresponding to a surge in recognition rate. The proposed method uses questionnaire-based recognition rates interpolated using Internet search volume, enabling trend period annotation on a weekly timescale. The main novelty is that we survey when the respondents recognize the entities that are highly likely to have trended and those that haven't. This procedure enables a balanced collection of both trending and non-trending entities. We constructed the dataset and verified its quality. We confirmed that the interests of entities estimated using Wikipedia information enables the efficient collection of trending entities a priori. We also confirmed that the Internet search volume agrees with public recognition rate among trending entities.

In this paper, we propose a rule discovery method that can reveal a combination of attributes that provide characteristic distribution of two consecutive variables of interest directly at high speed in a database having many attributes. In numerical association rule mining (NARM), when using association rules that handle consecutive numerical data values, it is difficult to heuristically extract rules that focus on statistical distributions of numerical data. The proposed method enables quick discovery of the number of rules necessary for prediction purposes using evolutionary calculations characterized by a network structure and a strategy to pool solutions throughout generations. This effectively finds attribute combinations in which the values taken by two consecutive variables of interest are both narrow ranges and can address instance-based two-dimensional regression problems in a short time. As an evaluation experiment, a prediction task using musical data linked with map data was carried out, and the discovery condition of the flexible rule was set. This resulted in realizing a high coverage rate in the instance-based regression problem, and the proposed method was effective in rule discovery based on the statistical distribution in NARM.

We propose a method to automatically classify eye-blink types using the eye-blink waveform integral value. The method is assumed to apply to an input interface using eye and It performs automatic detection of intentional blinks. Attempts to treat eye gestures and blinks as input channels in addition to conventional gaze input has studied due to the spread of gaze tracking and gaze input interfaces recently. However, classifying the eye-blink type as intentional or spontaneous using existing eye-blink classification methods is difficult because eye-blinks are highly individual motions that are significantly influenced by various conditions. Therefore, in this research, we construct a more robust measurement environment, which does not require a strict setting such as fixing the relative distance between the face and the camera even for non-contact measurement. In order to realize this, we defined new feature parameters are defined to correct the individual differences from moving image measuring by Web camera to assume applying on mobile interface. The proposed method performs automatic detection of intentional blinks by automatically determining the threshold of blink types based on the waveform integration value as new feature parameter. We also constructed a blink measurement system to evaluate the proposed method and evaluated the proposed method by experiment. The system splits the interlaced image field into disparate fields for blink measurement with sufficient temporal resolution. It then extracts the waveform feature parameters and automatically classifies the eye-blink types. Experimental results show successful classification of intentional eye-blinks with 86% average accuracy, thus demonstrated the high accuracy of the proposed method compared to conventional methods based on eye-blink duration.

In this study, brain-computer interfaces (BCIs) using visual evoked potentials (VEPs) induced by blinking light have been studied. We examined them to change the phase and frequency of the blinking light with variable onset intervals and changing luminance to realize multiple choices in the BCI based on code modulated visual evoked potentials (cVEP) using transient VEP. We used three types of blinking lights with fluctuation interval in addition to one blinking light with fixed interval and attempted to discriminate between the four blinking lights that a subject observed. The amplitude of averaged VEPs with fluctuation interval decreased when a subject did not observe the blinking light. The discrimination rate for the blinking lights was approximately 84%. In addition, we examined the changing luminance of the blinking stimulus to increase the number of multiple choices. In this study, we obtained VEPs based on synchronous addition method and verified the effects of changing luminance and lighting interval in indicators. In addition, our study aims to increase the number of simultaneous choices of cVEP-based BCI using blinking light with variable onset interval and changing luminance. In this paper, we report the possibility of multiple choices by conducting experiments for the estimation of proposed methods.

Reminders are used in various situations. The users of a location-based reminder system must specify the location to receive the notification, and spend extra time doing so. However, if the location can be estimated from the task that the user enters, this step can be eliminated. The authors focused on To-Do list, i.e., tasks the user wants to accomplish. It was assumed that the location where the To-Do can be accomplished could be estimated by parsing the To-Do. In this paper, the authors propose a method of estimating the place where the given To-Do can be accomplished. A list of daily To-Dos was collected from students through a survey questionnaire, and the To-Dos were used to evaluate the proposed method. In addition, the authors introduced a prototype of a location-based reminder system that applied the proposed method.