HHF Part5 - Active Haptic Sensing

Part 5 Active Haptic Sensing

08/18/2024

"Role of Hand Movements"

Exploratory Procedures (EPs)

EPs also varied in terms of their generality(sometimes called breadth of sufficiency),that is,in terms of the total number of object attributes that each EP could extract with above-chance (sufficient) performance (Lederman ' Klatzky, 1987). For example, lateral motion and pressure EPs are relatively less general because they provide only three different types of object information (i.e., texture, hardness, and thermal properties).

Material dimensions (e.g., rough/smooth, hard/soft, warm/cool) are generally processed more effectively than are geometric dimensions (e.g., relative position, relative orientation) (Klatzky & Lederman, 2003a, 2003c).

The differences in relative EP efficiency (as measured by accuracy and speed) led to the prediction that variations along material dimensions ~

Collectively, the results confirm that haptic observers choose to form object representations that are based more on material than on geometric properties when the perceptual differences within each dimension are perceptually equivalent across dimensions.

Recognition of Common Objects

Unimodal performance worsened while cross-modal performance improved when objects were rotated 180° about both the X (horizontal) and Y (vertical) axes between learning and test phases. Apparently, people recognize objects visually best from the front; in contrast, they haptically recognize objects best from the back, indicating that haptic and visual object representations are both view specific, but in strikingly different ways. Such a result further implies the need for some type of translation process. The cross-modal data also indicate that such intersensory processing can be relatively efficient.

 

"Material Properties"

Texture

We have previously noted that, provided there is relative motion between the skin and surface and that it is equated with respect to speed and applied force, roughness perception is equivalent whether the surface is moved over the stationary finger (passive tactile sensing) or the observer moves the finger over the stationary surface (active tactile sensing) (Lamb, 1983; Lederman, 1981). This implies that cutaneous mechanoreceptors provide the primary source of information for roughness perception.

Compliance

이건 뭐 다 연구 결과가 다름...: "suggesting that cutaneous inputs do not play a role"

When observers judge the stiffness (inverse of compliance) of springs enclosed in cylinders with rigid surfaces, a Weber fraction of about 17% isobtained. Skin anesthesia and joint anesthesia have no effect on stiffness thresholds (Roland & Ladegaard-Pedersen, 1977), suggesting that cutaneous inputs do not play a role. Other studies have obtained Weber fractions with active touch that range from 8% to 22% depending on the cues available in the particular task.

Thermal Properties, Material Composition 스킵

Weight

Size-weight illusion이 haptipc only 에서도 발생하기 때문에, haptic cue은 이 illusion (같은 무게 큰 부피가 더 가볍다고 인식되는 현상)의 필요충분 조건임. 

Researchers have demonstrated that observers use more than object mass to assess apparent weight. Haptic weight perception is strongly influenced by the object's size, which is a geometric property. This distortion of weight perception is known as the size-weight illusion: A large and small object of the same mass do not appear to be equally heavy; rather, the large object is judged to be lighter than the small one (Charpentier, 1891). The classic form of this illusion involves both visual and haptic inputs, that is, observers heft the object while looking at it. However, Ellis and Lederman (1993) have confirmed the existence of a purely haptic version of the illusion.

Once again, four material-exploration conditions were included: unimodal haptics (whole-hand grasp), unimodal vision, bimodal haptics and vision (traditional illusion), and a control condition that eliminated both haptic and material cues. For the low-mass objects in all but the control condition, the aluminum object was judged to be the lightest, followed in turn by the wood, then the styrofoam object (figure 5.6B). In keeping with the size-weight illusion, the magnitude of the illusion was equivalent for the traditional bimodal and haptics-only conditions, only moderate for the vision-only condition, and did not occur in the control condition. Thus, haptic cues are both sufficient and necessary, whereas visual cues are merely sufficient. The illusion did not occur with the high-mass objects. Ellis and Lederman proposed that the material-weight illusion was the result of using the cutaneous inputs to judge object material, which in turn influenced judgments of weight. As with the size-weight illusion, they attributed the material-weight illusion

"Geometric Properties"

Curvature

Curvarture detection threshold (discrimination between non-curvature and curvature threshold)가 가장 작은 mechanoreceptor보다 작음. 띠용. 그리고 lateral force랑 관련없는 줄 알았는데, 요즘 연구 (그래봤자 2003년 전)에서는 있다고 함. 

The discrimination thresholds for the narrow-width stimuli were close to the 1 μm threshold for detecting the presence of a single asperity against a smooth background previously noted by Johansson and LaMotte (1983). Note that this value is an order of magnitude smaller than the diameter of the cutaneous mechanoreceptors ( ̃10 μm), which lie approximately 0.8 mm below the skin surface at the interface between the epidermis and the dermis. As Louw et al. (2000) comment, this raises an interesting question concerning the nature of the mechanism that limits haptic resolution (cf. photons for visual thresholds and Brownian movement of the air at the eardrum for auditory thresholds).

Until recently, researchers have assumed that the haptic perception of curvature is determined solely by surface geometry as opposed to lateral force.

The results of this study highlight the fact that lateral force cues may well play a significant role in the haptic perception of curvature.

Orientation and Angle

The discrimination threshold for 75% correct performance was 5.2°. Voisin, Lamarre, and Chapman (2002) empirically confirmed that both cutaneous feedback from the fingers and kinesthetic feedback from the shoulders contributed to judgments in the previous task. Eliminating either source of sensory information raised angular thresholds, while eliminating both sources through passive touch with digital anesthesia prevented observers from performing the task at all.

As previously observed with tactile sensing (Lechelt, 1988), people are not equally sensitive to all line orientations (Appelle & Gravetter, 1985; Lechelt, Eliuk, & Tanne, 1976; Lederman & Taylor,1969). For example, Appelle and Gravetter (1985) and Lechelt et al. (1976) have confirmed haptic versions of the oblique effect, in which observers are less accurate when discriminating lines that are oriented obliquely than those oriented either vertically or horizontally.

Size

Visual과 Haptic sensing이 비슷한 illusion을 공유함. <->>-< 랑 T (horizonal - vertical line length)

The exponent for the psychophysical function describing haptically perceived length as a function of physical length is a power function with an exponent close to 1, indicating that unlike most other sensory judgments, there is excellent correspondence between perceived and physical length (Teghtsoonian & Teghtsoonian, 1965). 

In the Müller-Lyer illusion (figure 5.8), a line enclosed by fins is overestimated relative to the same physical length enclosed by arrowheads. A haptic variant of this illusion has been confirmed as well (Casla, Blanco, & Travieso, 1999; Millar & Al-Attar, 2002; Wong, 1975).

Recall another powerful illusion of length distortion that has been demonstrated both visually and haptically: the horizontal-vertical illusion (figure 4.18). A vertical line is consistently overestimated relative to a horizontal line of the same length. However, the distortion may be broken down into two separate components: a bisection effect and a radial-tangential effect. The bisection effect consists of the bisected line being underestimated relative to the bisecting line, as demonstrated with T-shaped figures. Similar bisection effects have been documented both visually and haptically. However, the radial-tangential effect (Heller & Joyner, 1993; Marchetti & Lederman, 1983; Wong, 1977), an anisotropy of space, only occurs haptically. Wong (1977) proposed that radial movements toward and away from the body are overestimated relative to those that are tangential to the body.

Shape of Two-Dimensional and Three-Dimensional Forms Larger Than the Fingertip

Temporal aspect를 고려해야 함.

Short segments tended to be overestimated, while long ones were underestimated. The same pattern was documented with acute and obtuse angles, respectively. Reproductions tended to be more regular overall than the actual stimulus shapes; however, the distortions in the reproduced shapes could be interpreted in terms of haptic distortions that we have previously discussed in the section “Size” (the radial-tangential and oblique effects). Interestingly, the accuracy in judging shape was about the same as when judging single lines or angles, implying that there is no net error introduced when several line segments are integrated into an overall shape.

Lakatos and Marks (1999) further suggest that salient small-scale features may initially draw the observer's attention away from the global structureuntilatimewhensuccessivehandmovementsservetoperformatemporallow-passspatialfiltering,therebydiminishingthe importance of the local features and, conversely, enhancing the significance of the global structure. As they remark: “The notion of touch as a modality that weights features differentially over time based on a continuously evolving impression of local and global shape is one that merits further attention” (p. 908)

 

Aristotle Illusion

The implication of the Aristotle Illusion is that tactile perception is not solely determined by the physical stimuli but also by the brain's interpretation of sensory information. It demonstrates that the brain uses contextual information, such as the usual positions and orientations of the fingers, to interpret tactile sensations. When this context is altered, as in the case of crossing fingers, the brain can be "tricked" into a misperception, revealing the complexity of how the brain processes sensory inputs to create our perception of touch.

"Attributes of Wielded Objects"

https://redstarhong.tistory.com/333

"Sensory Gating with Active Movements"

**0. Sensory Gating with Active Movements:

• Keywords:
  • Sensory Gating: The reduction of sensory signal transmission during active movement.
  • Dorsal Column-Medial Lemniscal Pathway: A neural pathway that carries sensory information from the skin to the brain.
• Takeaways:
  • Voluntary movements reduce sensory input transmission, affecting how we perceive touch stimuli.
  • Despite sensory gating, efficient active control during touch can maintain the accuracy of pattern recognition and perceived roughness.

 

"Sensing via Intermediate Links"

**0. Sensing via Intermediate Links:

• Keywords:
  • Active Haptic Perception: Exploring the environment through touch, often using tools or devices.
  • Intermediate Tools: Objects like pencils or haptic interfaces used to extend the reach of the hand in sensing.
• Takeaways:
  • Active haptic perception isn't limited to the bare hand; tools and devices can extend our sensory capabilities.
  • The complexity of these tools can range from simple objects to advanced haptic interfaces for teleoperation.

**1. Vibration Sensitivity:

• Keywords:
  • Vibration Sensitivity: The ability to detect vibrations, even of very small amplitudes, through the skin or a tool.
  • Vibrotactile Thresholds: The minimum amplitude of vibration detectable by a person, affected by factors like frequency, contact force, and stimulus location.
• Takeaways:
  • Humans can detect extremely subtle vibrations, sometimes at levels close to the physical limits of skin sensitivity.
  • Factors such as vibration direction, contact area, and the presence of a surrounding tool can influence vibration sensitivity, explaining variability in detection thresholds across different studies.

**2. Recognition of Common Objects:

• Keywords:
  • Haptic Exploration: The process of exploring objects through touch.
  • Rigid Probe/Rigid Sheath: Tools that limit direct tactile feedback during object exploration.
• Takeaways:
  • Recognition of objects is slower and less accurate when using a rigid probe or sheath compared to bare-finger touch.
  • Rigid interfaces reduce or eliminate important sensory cues that are crucial for accurate object identification.

**3. Surface Roughness:

• Keywords:
  • Roughness Perception: The ability to sense and differentiate textures.
• Takeaways:
  • Perception of surface roughness is moderately impaired when using a probe, with differences in the psychophysical functions.
  • The shape of these functions (quadratic) is influenced by probe motion and tip size, which affects how the probe interacts with surface elements.

**4. Compliance:

• Keywords:
  • Compliance Perception: Sensing the softness or hardness of materials.
  • Tapping vs. Pressing: Two methods of interacting with objects to assess their compliance.
• Takeaways:
  • Both tapping and pressing are effective for perceiving compliance, but tapping is generally more accurate.
  • Active sensing provides additional kinesthetic cues, leading to better performance in compliance tasks.

**5. Recognition and Discrimination of Object Material:

• Keywords:
  • Vibrotactile Cues: Sensory signals related to vibrations that help in identifying materials.
  • Material Recognition: The ability to identify materials based on their texture and other sensory cues.
  • Acoustic Cues: Sounds generated by contact, aiding in material identification.
• Takeaways:
  • Materials can often be identified by the vibrations and sounds produced when an intermediate link, like a probe, interacts with them.
  • Cutaneous and kinesthetic cues, along with auditory information, enhance the ability to discriminate between different materials.

**6. Shape and Size:

• Keywords:
  • Geometric Features: Characteristics like shape and size that are crucial for object identification.
  • Remote Touch: Using tools like a probe to sense objects, which affects speed and accuracy.
• Takeaways:
  • Judging material properties like roughness and compliance is possible with rigid tools, but identifying objects based on shape and size is more challenging.
  • Remote touch through a rigid link can slow down or reduce accuracy in recognizing objects, highlighting the limitations of indirect haptic perception.