PaperlayerView exact revision on arXiv

Source-linked AI summary

Externalizing Hand Grip Pressure Through Real-Time Visual Feedback

Andrew L. Huard, Avinash Ajit Nargund, Misha Sra

arXiv:2610.36769v1cs.HC

TL;DR

Grip pressure is difficult to observe and communicate, despite its importance in skilled grasping and the limits of existing instructional cues. The paper tests an instrumented glove with continuous and discrete visual feedback and grip-tailored mappings. Continuous feedback yielded more accurate matches and lower gaze entropy than bars, while tailored mappings reduced perceived effort and temporal demand.

  • Problem

    Instructional videos, demonstrations, and verbal cues provide limited real-time, spatially resolved feedback about where grip force should change and by how much,.

    Source: §I · p. 1

  • Method

    A mixed-design novice study (N=28) compared Gradient and Bar pressure displays and tailored versus generic four-region mappings using expert-recorded grip profiles.

    Source: §I · p. 2, C. Sensor-Region Mapping · p. 4

  • Results

    The continuous Gradient produced significantly more accurate pressure matching and lower gaze entropy than the Bar, while tailored mappings reduced perceived effort and temporal demand without improving objective matching accuracy.

    Source: §I · p. 2

  • Takeaways & Limitations

    Continuous pressure feedback supported more accurate matching and more concentrated viewing, while participants also assigned the Bar a role in precise magnitude guidance and Gradient a role in spatial guidance.

    Source: 1) “How Much” Versus “Where” · p. 10, §X · p. 13

    Because sport was treated as task context and analyses collapsed across baseball and table tennis, the study does not report sport-specific effects.

    Source: §VIII · p. 13

Abstract

from arXiv · show

Grip pressure, including contact force and its distribution across the hand, is central to skilled grasping but is not directly visible. This makes it difficult to teach, learn, and self-correct in grip-dependent motor tasks, where verbal instruction and tactile demonstration provide only partial guidance and no real-time, spatially resolved feedback tied to the learner's own grip. Using an instrumented glove, we compare pressure feedback visualizations for such guidance, evaluating how visualization type and, secondarily, sensor-region mapping shape pressure-matching performance in novices. In a study using baseball and table tennis as representative grip tasks (N=28), we find that a continuous gradient visualization produces significantly more accurate pressure matches and lower gaze entropy than a discrete bar display, despite participants perceiving the bar as more precise for pressure magnitude. Perceived precision and objective performance therefore diverged: the encoding participants judged more precise was not the one they matched more accurately with. Grip-tailored sensor-region mappings further reduced perceived temporal demand and effort, although this subjective workload advantage did not extend to objective matching accuracy. Overall, continuous pressure feedback supported more accurate grip matching and more concentrated viewing, and we present design considerations for real-time pressure feedback in motor skill guidance.

I. INTRODUCTION

Grip pressure is difficult to observe and communicate, while common instructional channels provide limited real-time, spatially resolved guidance. This work evaluates an instrumented glove and two visual encodings, finding more accurate matching with continuous feedback than discrete bars.

I. INTRODUCTION

  • Grip pressure encodes where and how force is distributed across the hand over time, but is rarely available as explicit feedback to learners or instructors.It is often inferred indirectly from posture, tension, or object movement.Source: §I · p. 1
  • Videos, books, demonstrations, and verbal cues provide only partial guidance, lacking real-time spatial detail about where force should change and by how much,.Concurrent visual cues may support in-task correction, but must remain interpretable within limited attentional capacity,,.Source: §I · p. 1
  • The system aggregates dense hand-pressure data into task-relevant regions and compares continuous Gradient and discrete Bar displays against expert-recorded grip profiles.The study uses baseball and table-tennis grips selected for recurring virtual-finger structures.Source: §I · p. 2
  • The continuous Gradient encoding produced significantly more accurate pressure matching and lower gaze entropy than the discrete Bar encoding in a novice study (N=28).Participants did not strongly prefer Gradient.Source: §I · p. 2
  • Grip-tailored sensor-region mappings reduced perceived effort and temporal demand, but did not significantly change objective matching accuracy versus generic mappings.The paper presents design implications for real-time visual feedback that externalizes tacit pressure information.Source: §I · p. 2

II. RELATED WORK

Prior motor-guidance systems commonly measure motion or muscle activity, while grip pressure distribution at the hand-object contact remains less directly observed. This work situates visual pressure feedback within the design tradeoffs of active, concurrent guidance.

A. Interactive Feedback for Motor Skill Guidance

  • Kinematic feedback compares novice pose or movement with an expert reference, but offers limited access to the internal and contact-level mechanics producing that motion –.Neuromuscular sensing extends feedback to muscle activity –, but neither approach directly captures force distribution at the hand-object contact surface.Source: A. Interactive Feedback for Motor… · pp. 2–3

B. Feedback Modalities in Motor Guidance

  • Physically enforced guidance can reduce early errors, but excessive or continuous assistance may foster scaffold dependency and limit transfer,,,,.Concurrent visual feedback instead requires users to interpret their state and intentionally correct their own movement,.Source: B. Feedback Modalities in Motor Guidance · p. 3

C. Visualizing Force for Closed-Loop Control

  • Visual feedback for active grasping must balance detailed information against attentional capacity while preserving spatial structure for deciding where pressure should change,, –.Bars support accurate magnitude readout, whereas continuous spatial fields can support holistic interpretation; placing feedback near its corresponding hand region may reduce mental remapping –.Source: C. Visualizing Force for Closed-Loop… · p. 3

III. SYSTEM DESIGN

The system captures contact force with a sensor glove and presents aggregated pressure in real time. Its design uses regional aggregation and grip-specific sensor mappings to retain task-relevant structure while managing control and attention demands.

A. Pilot Study

  • The pilot found regional aggregation best balanced usability and grip specificity: showing every sensor was unworkable, while a single whole-hand value was too coarse.Participants reported limited time to respond and higher effort in the unaggregated condition, without improved matching.Source: A. Pilot Study · p. 4
  • Observed biomechanical coupling between hand regions motivated allocating sensors into four task-relevant regions based on virtual-finger structure.Pressure changes in one region could alter readings elsewhere, including across finger and palm regions stabilizing the same object –.Source: A. Pilot Study · p. 4

B. Pressure Glove Design

  • The glove uses 20 force-sensitive resistors; an Arduino digitizes their signals for a Unity application that renders feedback at 30 FPS.The application receives sensor data over serial at 22 Hz.Source: B. Pressure Glove Design · p. 4

C. Sensor-Region Mapping

  • Both tailored and generic mappings display four regions, isolating sensor assignment from the number of regions shown.The tailored mapping follows each grip’s virtual-finger structure; the generic comparison shifts finger-group boundaries away from that task-specific organization.Source: C. Sensor-Region Mapping · p. 4
  • The tailored table-tennis mapping separates palm, middle-ring-pinky, index finger, and thumb, while the baseball mapping separates palm, thumb, index-middle, and ring-pinky regions.The mappings reflect the handshake grip’s independent index finger and the baseball pitch’s reliance on index and middle fingers.Source: C. Sensor-Region Mapping · p. 4

D. Visualizations

The visualizations encode aggregated sensor-region pressure either as a spatially aligned continuous color field or as discrete bars. The Gradient field interpolates color across each sensor-region using the calibrated pressure extrema and their mesh locations.

D. Visualizations

  • Gradient renders zonal pressure as a spatially aligned continuous color field, while Bar represents average zonal pressure as discrete bars on a 1–5 scale.The Bar display includes one bar for each of four sensor-regions.Source: D. Visualizations · p. 4

1) Gradient Visualization:

  • The Gradient pipeline uses calibrated pressure extrema and their UV coordinates to set a color range and direction, then interpolates color across each mesh sensor-region.The mesh is a rigged hand model rendered at fixed scale rather than resized to individual hand dimensions.Source: Figure 3 · p. 5, 1) Gradient Visualization · p. 5
  • When no stable spatial direction can be estimated, a sensor-region receives a uniform color corresponding to its pressure value.This fallback applies when the direction-vector magnitude squared is below 0.00001.Source: 1) Gradient Visualization · p. 5

2) Bar Visualization:

The Bar display summarizes normalized sensor readings by region, and the study logged pressure and gaze data for later analysis. Figure 4 shows the alternative four-region sensor mappings for each sport.

2) Bar Visualization:

  • Each Bar visualization displays one bar per sensor-region, with fill driven by average normalized pressure and a numerical annotation scaled from 1 to 5.The average is computed from bounded normalized readings of active sensors in the region.Source: 2) Bar Visualization · p. 5

E. Data Logging

  • The application logged calibration values and raw readings from all 20 sensors, while gaze was recorded separately and synchronized with pressure data for offline processing.Gaze was recorded using a webcam-based desktop eye tracker.Source: E. Data Logging · pp. 5–6
  • Figure 4 depicts two four-region mappings for baseball and two for table tennis, labeled Mapping A and Mapping B.Panels (a) and (b) show baseball; panels (c) and (d) show table tennis.Source: Figure 4 · p. 5

F. Target Pressure Profiles

Expert-recorded grip profiles supplied the pressure targets, which were rescaled to each participant’s grip strength and shown beside the participant’s live hand. The evaluation used counterbalanced blocks of grip-matching trials across sports and visualization conditions.

F. Target Pressure Profiles

  • Six expert-recorded profiles covered fastball, change-up, and tag-out grips in baseball and forehand, backhand, and slice grips in table tennis.Two experts each recorded three repetitions per grip; readings were normalized to expert MVC and averaged, then rescaled to participant grip strength.Source: F. Target Pressure Profiles · p. 6
  • Each static target profile appeared on a second hand mesh beside the participant’s live hand, using the active visualization encoding for direct comparison.Profiles were recorded by experts using the same physical objects and instrumented glove as in the study.Source: F. Target Pressure Profiles · p. 6

IV. USER EVALUATION

  • The user evaluation tested novices in a quiet laboratory using a desktop display, with the gloved arm stabilized on a table.The Unity visualization ran at 30 FPS and pressure telemetry updated at 22 Hz.Source: §IV · p. 6

A. Study Procedure and Task

  • Participants completed 36 trials in four counterbalanced blocks, matching one of three sport-specific grips during each 20-second trial under Gradient or Bar feedback.Each block paired one sport with one visualization and included nine randomized trials; participants completed NASA-TLX after each block.Source: A. Study Procedure and Task · p. 6

B. Participants

The study recruited 28 participants and measured grip matching, gaze behavior, workload, and visualization preference. Mixed-effects models evaluated visualization and sensor-region mapping while accounting for participant-level variability.

B. Participants

  • 28 participants aged 18–37 years took part, matching the power-analysis target for a medium effect size under the stated assumptions.All were right-handed and reported no color-vision deficiencies or hand-related conditions affecting grip performance.Source: B. Participants · pp. 6–7

C. Dependent Variables

  • Grip matching was assessed with normalized-pressure MAE and four match metrics, while gaze measures included entropy and horizontal and vertical dispersion.NASA-TLX responses and visualization preferences were also collected.Source: C. Dependent Variables · p. 7

V. RESULTS

  • Linear and generalized linear mixed models tested mapping, visualization, and their interaction as fixed effects, with participant ID as a random intercept.The analysis averaged metrics across the two blocks sharing a visualization before modeling.Source: §V · pp. 7–8, C. Dependent Variables · p. 7

A. Grip-Matching Performance

Gradient feedback yielded more accurate grip matches and lower gaze entropy than Bar feedback, while time to first match and most workload measures did not differ significantly. Mapping affected temporal demand and modified some visualization effects.

A. Grip-Matching Performance

  • Gradient produced lower average match MAE (0.256 vs. 0.263) and best match MAE (0.228 vs. 0.239) than Bar; participants also made more matches (6.13 vs. 5.73).The match-count advantage was significant only with Mapping A; time to first match did not differ significantly between visualization conditions.Source: A. Grip-Matching Performance · p. 8, Figure 7 · p. 9

B. Task Load & Preference

  • Mapping B participants reported higher temporal demand than Mapping A participants (55.9 vs. 37.9), while overall weighted TLX and other listed workload measures showed no significant effects.Effort differed by mapping only in the Bar condition: Mapping A reported lower effort than Mapping B (54.3 vs. 72.5). Visualization preference was evenly split, with 14 participants favoring each condition.Source: B. Task Load & Preference · p. 8

C. Gaze Metrics

  • Bar produced higher gaze entropy than Gradient (6.44 vs. 6.09 bits), while gaze dispersion was greater with Gradient on both horizontal and vertical axes.The entropy difference occurred under both mappings but was larger in Mapping A; mapping had no significant main effect on gaze dispersion.Source: C. Gaze Metrics · p. 9, Figure 8 · p. 9

VI. QUALITATIVE RESULTS

Participants assigned the Bar and Gradient different roles: the Bar communicated pressure magnitude, while the Gradient conveyed where to apply pressure. They also described different stability and workload trade-offs.

1) “How Much” Versus “Where”:

  • Participants described the Bar as better for judging how much pressure to apply and the Gradient as better for locating where pressure adjustments were needed.Ten participants found the Bar more effective for precise pressure intensity; participants also described the Gradient’s shading and layout as useful for locating hand regions.Source: 1) “How Much” Versus “Where” · p. 10

2) Perceived Stability and Noise:

  • Participants’ stability perceptions split: some found the Gradient noisy, while others experienced it as smoother; the Bar felt stable to some but its numerical targets could oscillate.Some participants found the Bar’s fluctuating values frustrating to hold at the target.Source: 2) Perceived Stability and Noise · p. 10

3) Attentional and Physical Demand:

  • Workload depended on the encoding: Gradient’s spatial detail could require monitoring across the hand, while Bar’s numerical targets could demand translation into physical adjustments.Some participants found Bar’s smaller set of cues more manageable, while others described it as more attentionally and physically demanding.Source: 3) Attentional and Physical Demand · p. 10

4) Strategies for Managing Multiple Regions Under Time Pressure:

The discussion links Gradient’s spatially aligned feedback to more accurate corrections, while participant accounts describe sequential, selective strategies and increasing reliance on intuition with familiarity.

4) Strategies for Managing Multiple Regions Under Time Pressure:

  • Participants managed the 20-second matching window by adjusting one region at a time, prioritizing key fingers, and often stopping at a “close enough” match.Some worried that further adjustment could disrupt a partial match elsewhere.Source: 4) Strategies for Managing Multiple… · p. 10

5) Shifting Reliance on the Display:

  • With familiarity, participants described moving from deliberate display reading toward instinctive, holistic adjustments and a developing sense of hand position and force.Some participants without prior sports experience reported recognizing proper gripping techniques for the first time.Source: 5) Shifting Reliance on the Display · p. 11

VII. DISCUSSION AND DESIGN IMPLICATIONS

  • The authors interpret the findings as indicating that visualization and sensor-region mapping jointly shape how participants process feedback and correct errors.They characterize the study as an exploratory investigation.Source: §VII · p. 11

A. Continuous visual encodings help with accurate corrections.

  • Gradient yielded more accurate matching, and preserving hand topography may reduce the translation from visual errors to motor corrections.The displays used the same 20 sensors aggregated into four regions; the authors recommend continuous spatial visualizations when precision is the primary objective.Source: A. Continuous visual encodings help… · p. 11

B. Visual encodings shape attention and correction strategies.

The discussion distinguishes the attentional demands of the encodings, the mismatch between perceived precision and matching accuracy, and the workload effects of sensor-region mapping. The authors recommend aligning visual and sensor-region structure with the grip task.

B. Visual encodings shape attention and correction strategies.

  • Bar’s narrower gaze footprint accompanied higher gaze entropy and matching error, whereas Gradient’s broader gaze dispersion accompanied lower entropy.The authors suggest explicit links between bars and hand regions may reduce correspondence-search costs.Source: B. Visual encodings shape attention… · pp. 11–12

C. Perceived magnitude precision did not track matching accuracy.

  • Participants judged Bar more precise for pressure magnitude even though Gradient produced more accurate matches, more matches, and lower gaze entropy.The authors attribute this divergence partly to the task rewarding reproduction of a pressure distribution across the hand rather than one region’s exact target magnitude.Source: C. Perceived magnitude precision did… · p. 12

D. Mapping alignment shapes the effectiveness of visual feedback.

  • Mapping affected reported workload but had no main effect on matching error, throughput, or gaze measures; Mapping B was associated with higher temporal demand.The Gradient match-count advantage appeared only under Mapping A, while the effort cost of Mapping B appeared only with Bar; the authors recommend grip-grounded mappings paired with continuous spatial encodings.Source: D. Mapping alignment shapes the… · p. 12

VIII. LIMITATIONS

The conclusion reports better matching and lower gaze entropy with Gradient, alongside reduced perceived temporal demand and effort from grip-tailored mappings. The authors bound these findings to a right-handed, fixed-form prototype, static laboratory tasks, and analyses collapsed across sports, and call for transfer and dynamic-setting studies.

VIII. LIMITATIONS

  • The prototype was designed only for right-handed users in one fixed form factor, and FSR noise artifacts can add cognitive load.The authors suggest higher-fidelity sensing hardware could potentially mitigate signal artifacts.Source: §VIII · pp. 12–13
  • Uncustomized glove fit and hand-morphology differences may shift sensor placement and contribute to MAE independently of reproduction accuracy.Reference profiles used the same off-the-shelf glove and carry sensor-placement variability too.Source: §VIII · p. 13
  • Static laboratory grip adjustments do not account for the biomechanical forces and attentional demands of dynamic sports movements.Source: §VIII · p. 13
  • Because analyses collapsed across sports, the study does not report sport-specific effects.Sport was treated as task context because mappings were defined by grip type and a sport–mapping interaction was expected.Source: §VIII · p. 13

IX. FUTURE WORK

  • The authors call for testing whether tacit grip skills are retained after real-time feedback is removed.Source: §IX · p. 13
  • Future designs could adaptively move from the accessible Bar encoding to Gradient once grip-matching MAE stabilizes below a threshold.Source: §IX · p. 13
  • The authors propose evaluating grips in real-world, high-velocity scenarios, requiring more robust, noise-resistant hardware that accommodates varied hand sizes and handedness.Source: §IX · p. 13

X. CONCLUSION

  • Gradient yielded more accurate matching and lower gaze entropy without measurable workload change, while grip-tailored mappings reduced perceived temporal demand and effort.Many participants found Bar more intuitive for initial familiarization.Source: §X · p. 13
Loading 2610.36769v1…