Vibrotactile display coding for a balance prosthesis.

Abstract

Preliminary experiments have demonstrated the potential usefulness of a precursor balance prosthesis that displays the tilt of the subject using tactile vibrators (tactors) which are in contact with the subject's skin. The device consists of a motion sensing system mounted on the head or body whose signals are converted into estimates of head or body tilt. Tilt is displayed to the subject by coding the tilt estimate into signals that are sent to the tactors using one of several schemes. Because full blown, end-to-end balance experiments are relatively time consuming and expensive, and because there are many possible display schemes, we have developed a quantitative means to evaluate the display step separately. We used a modified version of the manual control critical tracking task (CTT) to help us make an initial selection of the more promising vibrotactile display schemes for further evaluation. The classic CTT is a compensatory form of tracking in which the operator attempts to control an increasingly unstable system using a joystick to regulate a tracking error signal (system minus joystick) that is visually displayed as a dot on a cathode ray tube. Our modification added vibrotactile display of the error signal. For a given subject and vibrotactile display scheme, the level of difficulty at which the subjects lost control, called the critical lambda (lambda(c)), was highly repeatable. Four different coding schemes were evaluated using an array of 16 vibrators that were attached to the lower backs of 11 healthy subjects. The first scheme, called interval-based coding, modulated the interval between pulses that were sent to single tactors mounted on the subject's right and left side. A greater tracking error magnitude was displayed as a faster pulse rate. A positive error was displayed on the right side while a negative one was displayed on the left. The remaining three schemes, called position-based coding, used a horizontal row of 14 tactors. Tracking error magnitude was mapped to position of the activated vibrator so that an error near zero corresponded to a vibrator near the center of the back. The three position-based schemes tested used three, four, or seven tactors per side. Averaged over all subjects, the value of lambda(c) for the interval-based scheme was significantly less than it was for each of the three position-based schemes. There was no significant change in lambda(c) as the number of position-based tactors was increased from three to seven per side. The prediction of better actual balance performance using position-based relative to interval-based vibrotactile display was validated by a preliminary study of six normal subjects that compared the body sway produced during quiet standing while providing head tilt estimates using both display modes. Our study provides basic characterization using lambda(c) for several vibrotactile display schemes in human subjects. The quantitative CTT measure of performance can logically be extended to other applications of vibrotactile displays and to other kinds of display schemes used for rehabilitation.