• Home
• About SNAPI
• Related information
• What’s new
• Contact

Tactile Perception and Design

Introduction

Within this project, minimising the visual load is imperative to ensure distraction from driving is not instigated whilst using in-car controls.

"In a complete human -machine interface, the use of non-visual feedback modalities (such as auditory or tactile feedback) are expected to yield performance improvements in cases where the visual channel is near capacity" (Akamatsu, Mackenzie and Hasbrouc, 1995).

When vision is not suitable for a secondary task such as with an automobile navigation system, using other sensory modalities enables users to maintain their attention on the primary task whilst accurately completing other components of the task. Fleeting auditory feedback is problematic in an automobile situation especially when environmental noise is informative e.g. to warn of approaching police, the engine chugging, or where in-car noise e.g. stereo, talking, is predominant.

Akamatsu, Mackenzie and Hasbrouc (1995) whilst experimenting with a multi-modal mouse (with a tactile pin on the left button) concluded that tactile feedback produced a faster response time, "The addition of tactile stimuli yields quicker motor responses." This is probably attributable to the maintenance of the S-R compatibility because the stimulus is applied to a finger from which the response is delivered. This accords with previous research that visual reaction times are considerably slower than tactile reaction times in manual tasks (Boff and Lincoln, 1988). In light of this, a tactile basis would be more appropriate for this project.

It is thought that the constraints of the somatosensory system provide a foundation for the design specification. Initially, the tactual sense will be discussed before the ways users can interact with the devices, then various relevant devices will be commented on.

Tactual Sensing

The highest spatial resolution of the somatosensory system is found on the fingertips. So although experiments have been undertaken (Lechelt, 1984) on devices that deliver information through the skin around the body, the fingertips are regarded as the most accessible receptors. Touch perception is produced by spatiotemporal patterns upon the skin. More precisely, within the fingertips there are specific tactual sensors. These sensors can be categorised by: their size of active field and their response to static stimuli as in Figure 1.

Slow adapting (SA) sensors respond to perpendicular indentation of the skin and mainly static stimuli of a frequency below 30Hz. Rapid adapting (RA) sensors have no static response to stimuli but react to vibration. Type I units have small receptive fields whereas Type II units have larger receptive fields.

		Size of Active Field
		Type I	Type II
Response to static stimuli	SA	SAI	SAII
	RA	RAI	RAII

Figure 1: Summary of types of cutaneous sensors.

SAI, SAII and possibly RA too are excited by scanning a rough surface. RAI units register spatial information about skin deformation. They are in high density near the skin surface and are most easily excited by frequencies within the 10 - 60Hz range. They have a 3 - 4mm diameter receptive field. RAII units are located deeper within the tissue and are most easily excited by frequencies in the 60 - 1KHz range. These have large receptive fields of approximately 20mm in diameter so they do not inform of localised vibratory stimuli.

These sensors intricate responses are non-linear and time varying and their sensitivity varies with stimulus size, shape and duration (Kontarinis and Howe, 1995). So it is difficult to design for a particular situation. Despite this, a few generalisations are possible for example taken from frequency data by Johansson, Landstrom and Lundstrom (1982). The spatial resolution required for stimuli, is determined by the density of touch receptors in the skin. The resolution of peripheral mechanoreceptive units is approximately 1mm. Developing this Pawluk et al. (1998) inferred that 1cm X 1cm is the minimum physical size required for a tactile array.

Craig (1988) found that people with tactile experience on particular spatiotemporal patterns, showed improvements in identification of those patterns. Additionally, a general heightened tactual sensitivity affect was noted with tactually experienced people.

Regardless of the level of experience, Raj et al. (2001) list some limitations of the sense of touch.
- the limited bandwidth - the peak sensitivity is 250Hz.
- resolution - 2 point discrimination.
- susceptibility to habituation - brain disregards constant signals.
- adaptation - skin becomes less sensitive to constant stimuli.

Stevens (1992) reports that spatial acuity of the skin on the fingertips deteriorates with age. The average thresholds he determined are:
1.95mm for young people (18 - 33)
2.68mm for middle aged (41 - 63)
5.03mm for elderly (66 - 91)
The age range for drivers is likely to be nearly across this whole spectrum so attention needs to be paid to the older drivers and those with reduced tactile sensitivity.

A limit to the rapid processing of stimuli is due to masking. This arises when a secondary - masking - stimulus is presented in spatiotemporal contiguity to the target stimulus and causes deterioration of the target stimulus. Craig (1976) considered that interference by masking stimuli occurred at the same skin site either before (forward) or after (backward) the target letter had been presented. Backward masking interfering with letter recognition occured more than forward masking. Backward masking is prevalent for letters in which critical information for identification is located on the right side of the letter. When presenting letters in reverse, more forward masking for letters with critical information located on the left side of the letter resulted. Craig (1981) believes that masking is decreased by packaging brief letter presentations between longer pauses. So increasing the time between the target letter and the masking stimuli resulted in improved letter recognition. Weisenberger and Craig (1982) reported that subjects identified patterns in the absence of any masking stimuli and in the presence of spatially adjacent masking stimuli. The amount of interference in recognising the patterns, was measured as a function of the interval between target and masker onsets. The results indicated that more masking occurred when masker onsets followed target by 25 - 50 msecs than when onsets were simultaneous.

Passive / Active Touch

People can interact with tactile stimuli using passive touch (stationary) or active touch (scanning). Lederman, Loomis and Williams (1982) through experimentation found that cognitive processes appear to mainly take the SAI signals into account. So due to their localised nature, tactile structures can be read without any friction caused by movement (Fricke, 1994). So the display (stimulus) and finger (receptor) remain motionless.

This is contrary to various investigations into and manufacture of vibratory devices indicated shortly. For example Austin and Sleight (1952) found that with active reading outline figures were more distinguishable. Gill and James (1973) observed that the method of subjects inspection of tactile symbols differed. Some subjects placed their fingers directly on the symbols whereas others felt around the edges and in the centre of the symbol. The group using active touch performed better than those using passive touch.

Tactile Displays

The design of tactile displays should depend on their application and the users requirements. Tactile displays can be in a static, dynamic (moving) or of a vibrating form, each of these will be regarded below.

Static Displays

Static displays can be represented on embossed paper or solids materials.

A study by Gill and James (1973) classified tactile symbols into 3 types:
- Line symbols - to indicate boundaries of lines
- Areal / textural symbols - to indicate areas
- Point symbols - to indicate specific locators

Confusable point symbols as described by Jansson (1972) include:
- evenly embossed surfaces of different forms
- closed contours of different forms
- open contours of different forms
- combinations of similar units

Gill (1974) provides a starting place by indicating factors that influence tactile symbol discrimination. From this and subsequent studies a list can be derived.

- Elevation height
- Physical size
- Form (shape and texture)
- Orientation
- Stimulus redundancy
- Information content and association of meaning with symbol
- Configuration and information density
- Reference points
- Use of legends

Examining each of these factors in turn will reveal where there are still unexplored areas in our understanding.

Elevation Height

Distinguishability of areal, point and line symbols is easier when a differentiation between the symbol heights is adopted (Nolan and Morris, 1971). Furthermore, greater height demarcation enhances this. James and Gill (1975) whilst testing various symbols used a relief of 0.7mm.

Loomis (1985) examined the accuracy of tactile recognition of raised characters which he found to be a function of 5 variables:
- height of upper surface related to the surrounding base
- character set (uppercase roman and braille)
- spatial extent of the characters (2 sizes)
- mode of touch (static and moving)
- contact force (2 values) which gave no significant main effects.
Analysing the interactions indicated that only height by mode was significant whereas a moving finger permits a considerable improvement in performance over static touch for very shallow characters. A moving finger improves performance only slightly over static touch for higher characters.

Physical Size

Due to the inadequacy of the touch system, the physical size must be larger than if the visual system was being used to process the information. Nolan and Morris (1971) believe this needs to be at least 5mm. For static touch 2.3mm is specified by Boring (1942) and Weinstein (1968). James and Gill (1975) used areal symbols that were 50mm x 50mm and linear symbols that were 100mm in length to test their symbols.

Form

Major (1898) tested the distinguishability of people's tactile sense of solid and outline circles and triangles. He found outline circles were the easiest to discriminate and solid circles were the most difficult. Zigger and Barrett (1927) also discovered that outline shapes were more accurately distinguished than solid or punctate symbols. From this it can be inferred that when using the fingertip, an outline shape is more easily perceived possibly because there are two edges to feel as shown in figure 2. But at smaller sizes, the double edge may be a source of confusion if it is perceived as an areal pattern.

Figure 2

James and Gill (1975) found that for interrupted lines spacing is a distinguishing feature for dotted lines e.g.
…………
. . . . . . .

but not for dashed lines e.g.
----------
- - - - - -

lines with edges broken by vertical projections can be easily confused e.g.

Discriminability is paramount since James and Gill (1975) explain subjects spend longer discriminating symbols that are similar than those that are radically different. Subjects may be looking for subtle differences that do not exist.

Texture can provide further cues to aid identification of a symbol. Kops and Gardner (1996) defined texture by the
- intensity - dot density
- spatial cues - spacing
- angular orientation
Their findings include the relationships between these variables. Textures are distinctive when their elements are tightly spaced along one axis but widely spaced along other axes. High and low density patterns are easier to distinguish from each other than those with medium density textures. Orientation and density are easier to differentiate than density alone. Diagonal and diamond patterns are confused with each other especially with higher intensities. Vertical arrays are only recognisable at high densities whereas other orientations are evident at low and medium densities. There is an interaction between spacing and orientation so patterns that differ in both variables, are discriminated better than those that differ in only one variable. Patterns whose spacing and orientation are easy to visualise are easier to differentiate. Spatial cues are limited by the diameter of the cutaneous receptive fields. Kops and Gardner (1996) inferred that subjects seem to use spatial rather than intensity cues when discriminating patterns of similar intensity.

Direction can also be indicated by the texture of the symbols e.g. Schiff, Kaufer and Mosak (1966) developed a tactual line saw-tooth in cross-section which in one direction felt smooth whereas the other direction was rough.

Orientation

Goodnow (1969) reports his subjects found it easier to percieve a change in orientation than a change in shape. Lechelt (1985) found orientation deviation for patterns from horizontal and vertical positions rather than diagonal ones were more accurately discriminated. This was reiterated by Kops and Gardner (1996) who state vertical and horizontal orientations are more accurately recognised than oblique ones possibly because of the interaction when moving a digit across the pattern.

Stimulus Redundancy

At small sizes, increasing the stimulus redundancy incurs fewer errors (Schiff & Isikow, 1966). Regardless of the system implemented, Coules and Avery (1965) conclude that for fine-grain data discrimination, individual differences may influence the final design of the sensor. The differences can be reduced and the sensitivity of the user improved if the electronic design and transducers provide redundancy for the human. Cholewiak and Collins (1997) whilst measuring students susceptibility to interference in masking and discrimination tasks and their ability to identify vibrotactile patterns, found that individual differences were constant across tasks so abilities in a variety of pattern perception tasks might be predictable.

Information Content and Association of Meaning with Symbols

Caution must be applied when visually impaired people are used to specify a design for the sighted population because they may not have the same representation of the world and it's symbols. Totally blind people have difficulty perceiving a tactual version of the visual arrow. An exploration of semiology would deepen the understanding of the meaning of symbols.

Configuration and Information Density

Touch inspection differs from visual inspection in that it requires mechanical interaction with the inspected object. Although the display is unlikely to be read using the same perceptual and cognitive resources as their traditional equivalents, the basis of the users interaction (especially novices) will be from their stereotypes of reading on paper. Tactile displays are read from left to right and top to bottom. To ensure the congruency of the systems configuration and users interaction with it, findings by Gill (1974) indicate that performance can be improved if users are taught to scan in a systematic manner. Albeit, this would require some training or instruction.

Identification of points is superior under conditions of maximum symbol separation.
[See 'Texture' subheading above]

Reference points

Reference points provide a starting / stabilising place from which other functions can easily be found.

Use of legends

Nolan and Morris (1971) determined that the number of tactile areal or line symbols which are not distinct in a set may not exceed 8 or 10, this was also found by James and Gill (1975). They attribute this perceptual limit to the parameters which distinguish tactile symbols. Increasing these parameters may therefore increase the number of legible tactile symbols in a set. Although the number of legible tactile symbols in a set may be able to be increased by expanding these parameters but is this necessary? The more symbols that are included in the design, the greater the cognitive load imposed on the user. Would the inclusion of a legend be a suitable proposition? If the symbols are instinctive, this would be superfluous. Gill and James (1973) point out that symbols in context may be easier to identify although taken out of context this aspect is rendered ineffective.

Dynamic Displays

Craig (1981) reports that static displays produce better letter recognition. Although proprioceptive signals generated by movement are needed for optimal discrimination of large figures, Loo et al. (1983) found that active, passive and proprioceptive modes allowed better identification than movement of the shape. When the display moves, the finger is required during the whole transmission of the message. It is also easier to lose one's place in such an arrangement especially if other tasks are performed in parallel. For this investigation, passive displays using passive or active touch are deemed more suitable.

Vibrotactile Displays

Vibration can be thought of as the rate of skin deformation. Horner's (1992) results describe that there are limited attentional resources for processing vibrotactile patterns. By increasing understanding of the relation between vibration and task type, it can be ascertained which applications would benefit from this thereby improving design.

For very low frequencies the soft subcutaneous tissue under the contact areas of the fingers act as mechanical low pass filters which attenuate small amplitude vibrations before they reach receptors beneath the skin. For low frequencies, an array display may be appropriate to provide spatially resolved stimuli for the RAI and SA receptors (Cohn, et al., 1992; Hasser & Weisenberger, 1993; Kontarinis & Howe, 1993). High frequency vibrations are poorly localised on the skin indicating that a single vibratory display for each finger is adequate.

The amplitude and frequency of different devices varies. Developers of the Exeter array (2002) report that patterns of low frequency stimulation (40Hz) produce similar 'real' touch sensations than patterns at higher frequency of 320Hz. The range of motion used by Kontarinis and Howe (1995) was an amplitude of 3mm. They mention that the displays produce a peak force of 0.25N at 250Hz. This is reiterated by Raj et al. (2001) and Audiological Enginnering (2002) who report the peak frequency of the skin stimulator is 250Hz, which is quote to be 'the most sensitive frequency for the skin'.

Kontarinis and Howe (1995) describe that vibratory feedback on manipulation and inspection tasks enabled subjects to persistently complete the tasks successfully. This result varies for different tasks but for the reported experiments, the vibratory group performed the tasks significantly better. The combination of vibration and force feedback compliment each other and can engender another significant improvement in performance. When the mechanical state was judged in a manipulation task, vibratory information enhanced performance by decreasing reaction time. They concluded that the role of vibration is in indicating transitions (a sequence of phases or subtasks separated by discrete events) in complex tasks. These transitions are best detected by RAII. To conclude, Kontarinis and Howe (1995) say that "high frequency information may be effectively conveyed by specialised sensors or by displays that have good response at high frequency and minimal response at low frequency".

In terms of vibrotactile devices, Bliss and colleagues in the 1960s developed the Optacon as a reading device which enables visually impaired people to read printed material. A cigarette lighter sized camera is placed over the print to act as an optical pick-up which is translated onto a corresponding vibrotactile representation. This is achieved by means of vibrating metal pins consisting of 144 piezoelectric bimorph pins in a 24 x 6 matrix comprising an area of 13mm by 28mm. The piezoelectric bimorphs vibrate in a bi-level vibratory pattern at a fixed frequency of approximately 230Hz causing pins to impact the skin in a nonlinear manner. The index finger tip is placed over this display to 'read' the vibratory representation of the print.

Heller, Rogers and Perry (1990) reveal that tests with an Optacon indicate that for blindfolded subjects, number recognition was superior with active touch and the left hand. This could relate to the right hemispheres specialisation for tactile pattern detection and number recognition may involve minimal linguistic demands.

The Cutaneous Communication Laboratory at Princeton University has performed research into how information can be best transmitted through the skin. One of the experiments by Cholewiak and Collins (1990) found that a plastic membrane placed over the Optacon array did not interfere with vibrotactile pattern acuity or identification.

In addition to Telesensory and Canon's Optacon I and II, other vibrotactile devices and technology are mentioned below.

3D tactile feedback has also been investigated by Zimmerman et al. (1987) amongst others. 3D Touch Technology (Sensable, 2002) allows users to directly interactive with digital objects as they would in the real world.

Virtual tactile feedback has also been explored but is not thought suitable within an automobile context.

Bliss et al. (1970) tested a second optical to tactile image conversion system that allows information to be accessed from the environment as well. The single system has two separate sets of optics: One for printed information, and one for environmental sensing.

The 2D 'rabbit' display reported by Tan and Pentland (2001) has many features to display directional information in applications for example as a navigational device; it has been demonstrated as useful to provide simple directional cues to a driver.

Experimentation Needed

This report indicates that a static or vibrating display would be the most appropriate for the proposed car navigation system project. Active touch appears to provide the most information and can be induced by making symbols larger than the fingertip receptors so the user has to move their digit to decipher the whole symbol. This is in keeping with the fact that tactual perception requires larger symbols than if visual perception is being used to identify them. But what is the optimum size for these symbols? An experiment could be performed to determine this.

If a vibrating display is used, a frequency of 250Hz is supported by previous research, but does this affect the size of the symbols or other variables? If different heights are used for line symbols and point symbols, what are the heights that they can be manifested at whilst still being distinct from each other.

With a navigation system, depending on the format, direction would be an important cue. Further methods of showing direction could be found by testing different forms and patterns to decipher how users instinctively react to their portrayal.

In context, symbols can be more easily determined but they still need to be distinct from each other. These tests will need to be performed with sighted drivers due to the semiology of an automobile environment. Despite this the textures of the symbols can be tested with visually impaired subjects. The spacing is a more informative variable than the intensity of the pattern so this would need to be examined.

Reference points are regarded as advantageous but how are these best represented and induced to aid the task? Data from visually impaired users would enlighten this issue. These factors can be tested experimentally with visually impaired users to inform the design of an automobile navigation system.

 

Contents Page

References Page

 

• Join email list
• Privacy policy
• Links