Shinichi Hikita

Point-and-Click Interface Based on Parameter-Free Eye Tracking Technique Using a Single Camera

We propose a method for the estimation of point of gaze with neither user- nor environment-dependent parameters. The gaze direction is calculated from the centers of both the pupil and the eye rotation. The center of the eye rotation is determined using the centers of both the pupil and the iris and the edge of the iris when at least four calibration targets, for which only the distances between them are known, are fixated on the screen. The mean horizontal and vertical errors for seven subjects were 0.91 deg and 0.77 deg, respectively. Next, a point-and-click interface, in which a user can move a cursor by a gaze shift and click a computer mouse by a voluntary eye blink or short fixation, was developed. On average, it took 1.2, 0.9, and 0.8 sec to point and click for each target with eye blink, short fixation, and normal hand manipulation, respectively.

Modeling of patient's blood pressure variation during ambulance transportation

In an emergency transportation by ambulance, a patient is transported in a supine position. In this position, a patients blood pressure (BP) variation depending on an inertial force which occurs when an ambulance accelerates or decelerates. This BP variation causes a critical damage for a patent with brain disorder. In order to keep a patient stable during transportation, it is required to maintain small BP variation. To analyze the BP variation during transportation, a model of the BP variation has so far been made. But, it can estimate the BP variation only in braking. The purpose of this paper is to make a dynamical model of the BP variation which can simulate it in both braking and accelerating. First, to obtain the data to construct the model, we used a tilting bed to measure a head-to-foot acceleration and BP of fingertip. Based on this data, we build a mathematical model whose input is the head-to-foot acceleration and output is the Mean BP variation. It is a switched model which switches two models depending on the jerk. We add baroreceptor reflex to the model as a offset value.

Robust Critical Control for Servo Systems using Disturbance Observers

This paper develops a computationally efficient method to design a controller for servo systems which require that tracking errors and control inputs are strictly maintained within pre-specified constant bounds. To ensure these bounds in the presence of unknown disturbances and parameter variations of a plant, the disturbance observer is introduced. In the framework of the principle of matching, a controller is designed as a 2-degree-of-freedom controller including the disturbance observer under the condition that reference signals and disturbances are restricted in rate of change and magnitude, respectively. Through a case study of servo controller design for the actively-controlled bed for ambulances, it is shown that the 2-degree-of-freedom controller can compensate wider parameter variations compared to a 1-degree-of-freedom controller and that the proposed method is computationally efficient

Contribution of vestibular (otolith) and pursuit system to stabilizing vision during sinusoidal head movement

The purpose of this study was to investigate how the vestibular (otolith) system and the visual pursuit system contribute to stabilizing vision during head translation. Eye velocity when fixating a stationary target with sinusoidal head movement (visually enhanced LVOR, VLVOR) was compared to that of smooth pursuit (SP) with the same relative target velocity (20 to 180 deg/s) as in the case of VLVOR. The eye velocity during VLVOR was much higher than that during SP with increasing relative target velocity. The gain of the SP (eye velocity/target velocity) markedly decreased with increasing stimulus frequency (0.6 1.2 Hz) while the gain of the VLVOR decreased only slightly. The eye velocity during VLVOR depended on the target velocity, not on the velocity of head motion, and was not associated with the state of vergence. These findings suggest that the command signal of the VLVOR is generated mostly by the visual pursuit system, and that the signal from the otolith system modifies the command signal of the pursuit system to improve the target tracking performance.