James C. Bliss

A Relatively High-Resolution Reading Aid for the Blind

Many direct-translation reading aids for the blind have been built in the past, employing an auditory output consisting of a combination of tones indicating the black regions in a vertical slice through a letter space, or a tactile output consisting of a raised or vibrating image of the letter shapes. Maximum reading rates obtained by the majority of subjects with these reading aids have been less than 10 correct words per minute, and the cause of this limitation has not been well understood. By analyzing the spatial spectral content of letter patterns, we show here that most of these reading aids have violated the well-known sampling theorem. It is suggested that this may be a significant factor in the observed reading-rate limitation. The design of a reading aid is described based on the conclusions from this analysis of the sampling process, and on recent results from tactile research. With this reading aid a hand-held probe images a vertical section of a letter space onto a 24 × 6 array of photosensors, and the probe is manually moved horizontally across the line of print. The signal from each photosensor controls a tactile stimulator in a corresponding array of 144 stimulators, which are placed on a single finger. In preliminary reading tests with this device, four subjects have all read at rates greater than 10 correct words per minute, and two of the subjects have read at rates greater than 20 correct words per minute.

Information available In brief tactile presentations

Two experiments investigated characteristics of Immediate recall for brief tactile stimuli applied to the 24 interjoint regions of the fingers of both hands (thumbs excluded) The obtained Immediate-memory span varied from 35 to 7,5 stimulus positions correct after correction for guessing, similar to the results In analogous visual studies, Properties of any hypothetical tactile short-term memory were studied by requiring subjects to report only a specified portion of the stimuli presented, and by varying the time of occurrence of the marker specifying which portion of the stimuli to report, In this partial-report condition, subjects had more stimulus information available at the time of reporting than their immediate memory spans indicated, provided that the stimulus marker occurred within 0.8 sec after stimulus termination, The data suggest that at least for the amount of training employed here, any tactile short-term memory has much less capacity than an analogous visual short-term memory.

Perception of sequentially presented tactile point stimuli

Either two or three brief (10 msec) airjet stimuli were sequentially presented to any of the 24 interjoint regions of the fingers (thumbs excluded). The stimulus onset interval (SOI) ranged from zero (simultaneous presentation) through 200 msec. The S’s task in one part of the experiment was to report the positions stimulated in the order that the stimuli were presented; in a second part it was to rate the apparent motion produced by the stimulus sequence. While the ability of Ss to spatially localize the stimuli was a constant independent of SOI, their ability to temporally order the stimuli depended strongly on SOI. With two stimuli, these sequential errors decayed exponentially with SOI with a time constant of 26 msec. With three stimuli, however, both the sequential errors and equivalent temporal Urnen were more than twice as large as with two stimuli, indicating that the three-stimulus task is considerably more difficult than the two, and that the same simple temporal resolution model does not explain both cases. A model with a constant rate of information uptake, however, can explain both of these cases.

Tactile perception of sequentially presented spatial patterns

Tactile pattern recognition was studied by presenting pairs of alphabetic shapes in rapid succession at the same anatomical location, the subject being required on each trial to identify bath of the patterns. Experimental variables were the duration of each stimulus and the time between stimuli. Three aspects of the observed interaction were (1) an increase in letter reversals for very short interstimulus intervals; (2) a greater percentage of first-response errors for short-stimulus onset intervals and a greater percentage of second-response errors for long-stimulus onset intervals; and (3) a crossover in the first- and second-response error rates in the range of 100 to 200 msec. after the onset of the first stimulus. These results are consistent with some of the temporal properties of models proposed for analogous visual tasks.

Effect of display movement on tactile pattern perception

The effect of display movement on the ability of subjects to recognize alphabetic shapes tactually was investigated. The display consisted of a computer-controlled 8-by-6 array of small airjet stimulators that could be physically translated in a small circle by means of a mechanical linkage. The experimental parameters were the stimulus duration, the angular velocity of the display, and the amplitude of the rotation. Recognition accuracy increased with stimulus duration between 100 and 400 msec. For a rotation amplitude of 0.8 cm, a maximum in recognition accuracy occurred at a rotation velocity of 400 rpm, or 150 msec. per revolution. The optimum angular velocity appeared to decrease as the amplitude of rotation increased. From these results and certain related neurophysiological evidence, a hypothetical model is suggested which qualitatively can account for the data.

Optical Detector for Objects within an Adjustable Range

This paper describes a relatively simple system for detecting in-focus objects. A nonlinear, wide-area photocell with parallel-plate electrodes is vibrated axially behind a lens. When the object is out of focus, the photocell conductance varies at the vibration frequency. When the object is in focus, the photocell conductance signal “folds over” because of the symmetry about the image plane. This “fold over” is detected by electronic circuitry. The scheme is shown to operate over a wide field of view, and the photocell conductance as a function of axial position is given for a number of idealized object patterns. An experimental unit built on these principles is described briefly.

An information processing model for tactile perception

Sensory information processing model for tactile perception using array of airjet and piezoelectric stimulators applicable to display design and nervous system investigation