High-Bandwidth Biomimetic Finger for Tactile-Transparent Remote Texture Sensing

• Prior work links texture discrimination to spectral energy balance and identifies fingerprint ridges and layered mechanical gradients as ways to modulate texture-induced vibrations.Source: §1 · p. 1
• Many flexible sensors have upper frequency limits of 100–200 Hz, while fingertip modes span 100–260 Hz; how fingertip design affects spectral capture and perceptual transparency remains underexplored.Existing biomimetic sensors have often been optimized for texture discrimination accuracy.Source: §1 · p. 1
• The study examines how fingertip structure, geometry, and material properties affect vibrotactile signal spectra and, in turn, the tactile transparency of feedback.Three prototypes are compared with the human fingertip using temporal and spectral analyses and a user study.Source: §1 · p. 2

2.1 Structural Design of Biomimetic Finger

• Hard fingerprint ridges over soft substrates convert tangential friction into normal vibration, while ridge geometry and multilayer structures modulate texture signals.Prior work has established structural amplification, but its spectral and perceptual effects remain less explored.Source: §2.1 · p. 2

2.2 Sensing Mechanisms of Biomimetic Finger

• Spatial sensors capture contact patterns, while high-bandwidth temporal sensors capture vibration spectra; despite bandwidth advances, how robotic tactile signals mirror human perception remains underexplored.Reported temporal sensing bandwidths extend from roughly 250 Hz to 1000 Hz.Source: §2.2 · p. 3

2.3 Evaluation Metrics of Biomimetic Finger

• Prior evaluation emphasizes texture discrimination, but discrimination alone may not reflect perceptual authenticity, motivating waveform, spectral, and human-perception measures.The reviewed work describes three performance levels: enhanced signal capture, higher spatial resolution, and surpassing human performance on selected metrics.Source: §2.3 · pp. 3–4

3.1 Mechanical Design

• The design combines a skin layer, subcutaneous tissue, rigid bone scaffold, and centrally embedded IMU, with material choices spanning graded stiffness.The skin is 0.5 mm thick; PDMS or TPU-95A provides skin, and PDMS or Ecoflex-30 provides subcutaneous tissue.Source: §3.1 · p. 4

3.2 Fingertip and Textured Surfaces

• Three variants share a PETG scaffold and geometry; the molded PDMS designs have human-comparable ridge morphology, while the printed TPU-95A design has lower ridge fidelity.The eleven test surfaces include six periodic or random PLA surfaces and five natural surfaces.Source: §3.2 · pp. 5–6, Figure 2 · p. 5

3.3 Fabrication

• Fabrication assembles the bone and IMU with PDMS or Ecoflex-30, then bonds either molded PDMS skin or directly printed TPU-95A skin.The fabrication schematic distinguishes the molded PDMS and 3D-printed TPU-95A skin routes.Source: §3.3 · p. 6, Figure 3 · p. 6

3.4 Data Preprocessing

• Signals are demeaned and low-pass filtered at 300 Hz, then normalized differently for similarity analysis, visualization, and feedback rendering.Feedback signals are group-RMS aligned, divided by five times that RMS value, and clipped to [-1, +1].Source: §3.4 · p. 7

4.1 Data Acquisition with Fingers

• The robot arm translated each fingertip 40 mm at 10, 20, and 50 mm/s across 11 textured surfaces; three repetitions for four fingers produced 396 trials.The human index finger served as the reference, and both fingers were pressed 0.5 mm into each surface during sliding.Source: §4 · p. 7, §4.1 · p. 7

4 Evaluation Experiment

• The study compared biomimetic and human fingertip signals in temporal and spectral domains to identify design-dependent differences in signal fidelity.Those signal analyses informed the subsequent user study.Source: §4 · p. 7

4.2 Temporal Analysis of Vibrotactile Signals

• PDMS had higher overall temporal similarity at 20 and 50 mm/s (0.189 and 0.281), while PDMS+Ecoflex ranked highest at 10 mm/s (0.210).PDMS was most similar on random surfaces; PDMS+Ecoflex led on periodic surfaces at 10 mm/s and on metal surfaces at all speeds.Source: Figure 4 · p. 8, §4.2 · p. 8, Table 1 · p. 8
• The multilayer TPU+Ecoflex and PDMS+Ecoflex designs generated more intense vibrations than single-material PDMS, while periodic waveform signatures appeared for PDMS+Ecoflex and the human fingertip.On random surfaces, TPU+Ecoflex responded more strongly than PDMS; PDMS+Ecoflex signal energy followed roughness variation across random and natural surfaces.Source: §4.2 · p. 9

4.3 Spectral Features Analysis

• PDMS+Ecoflex had the highest mean characteristic-frequency peak prominence on periodic surfaces (f0 prominence 3.93), while the multilayer design amplified texture-induced vibrations across a wide frequency range.PDMS had the lowest average f0 prominence and band energy; PDMS+Ecoflex also produced pronounced peaks in the 150–300 Hz range resembling the human fingertip.Source: §4.3 · p. 10, Table 2 · p. 10
• PDMS+Ecoflex vibration energy rose with random-surface RMS height, whereas TPU+Ecoflex had severalfold greater 150–300 Hz band energy without systematic dependence on RMS.Across natural surfaces, M7 and M8 had comparable spectral features across fingertips; M9 produced prominent high-frequency energy except for PDMS, and both metal surfaces were dominated by DC side-lobes.Source: §4.3 · p. 10

4.4 Spectral Similarity Analysis

• PDMS led mean similarity in 0–150 Hz at all speeds, while PDMS+Ecoflex led in 150–300 Hz at 20 and 50 mm/s, reaching 0.73 versus 0.65 (PDMS) and 0.66 (TPU+Ecoflex) at 50 mm/s.At 50 mm/s on periodic surfaces, all three designs showed good similarity: 0.80 (PDMS+Ecoflex), 0.72 (PDMS), and 0.69 (TPU+Ecoflex).Source: §4.4 · pp. 10–11, Figure 6 · p. 11

5 User Study

• The user study assessed tactile transparency through perceptual agreement, complementing earlier comparisons of temporal and spectral signal characteristics.The study connected fingertip design with perceptual characteristics.Source: §5 · p. 11

5.2 Evaluation of Tactile Transparency

• PDMS+Ecoflex led in Roughness consistency (6/11 surfaces), TPU+Ecoflex in Repetitive (7/11), and PDMS in Granularity (7/11).The TPU+Ecoflex Repetitive advantage arose mainly on non-periodic surfaces (7/8 matches); PDMS was closest to bare-hand perception on periodic surfaces.Source: Figure 7 · p. 12, §5.2 · p. 13

5.1 Experimental Setup

• Thirty participants rated direct exploration and motor-delivered feedback on Roughness, Repetitiveness, and Granularity; feedback reproduced acceleration signals recorded at 20 mm/s.Each participant explored 11 surfaces directly and then rated feedback for 33 fingertip-surface combinations; paired ratings were compared with Wilcoxon signed-rank tests.Source: §5.1 · p. 12

5.2 Evaluation of Tactile Transparency

• On periodic surfaces, feedback conveyed weaker Repetitive perception than bare-hand exploration, with PDMS ratings aligning most closely with direct touch.On random surfaces, PDMS+Ecoflex Roughness ratings increased with RMS height; PDMS Granularity ratings matched bare-hand perception, while TPU+Ecoflex elicited pronounced Roughness and Granularity.Source: §5.2 · pp. 12–13
• On natural surfaces, feedback generally agreed with bare-hand perception; all fingertips matched Repetitive ratings for Oxford cloth and leather without significant differences (p > 0.05).On plant fiber, all designs matched bare-hand ratings at p > 0.01 across dimensions except TPU+Ecoflex in Roughness (p = 0.006).Source: §5.2 · p. 13
• On the smooth metal plate, all three fingertips differed from bare-hand ratings in Roughness and Granularity (p < 0.001), whereas Roughness ratings on the pebbled metal plate were comparable (p > 0.05).On the pebbled plate, PDMS differed marginally in Repetitive (p = 0.033), and PDMS+Ecoflex in Granularity (p = 0.011).Source: §5.2 · p. 13

6 Discussion

• The multilayer PDMS+Ecoflex design improves vibration intensity and texture-feature representation over single-material PDMS, while stiff TPU+Ecoflex skin boosts intensity but masks characteristic peaks.TPU+Ecoflex also elicits consistently high Roughness and Granularity ratings with poor discrimination across random surfaces.Source: §6 · pp. 13–14
• Signal similarity did not guarantee perceptual transparency: PDMS signals resembled the human fingertip yet conveyed less comparable roughness information, and periodicity perception was weaker across all three designs.On smooth metal M10, spectra similar to the human fingertip nevertheless evoked texture sensations absent in direct contact; residual robot-motion side-lobes are a plausible source.Source: §6 · p. 14
• Roughness draws on signal-energy variation with surface roughness, Granularity on overall spectral shape, and Repetitive perception on faithful representation of characteristic peaks.Source: §6 · p. 14
• The system is limited by single-contact feedback, vibration-only rendering, and IMU-captured system noise, motivating spatial, multimodal, and signal-conditioning improvements.The single contact point omits distributed mechanoreceptive cues; vibration acceleration excludes force, temperature, and skin-stretch cues.Source: §6 · p. 14

7 Conclusion

• The paper presents a human-inspired fingertip with multilayer mechanical gradient, fingerprint morphology, and an embedded IMU for capturing texture-induced vibrations for remote vibrotactile rendering.Source: §7 · p. 14
• The multilayer gradient most clearly improves signal intensity and texture-feature representation, whereas a stiffer skin layer amplifies vibration at the expense of feature representation.Source: §7 · p. 14
• The remaining gaps are in acquisition and rendering, with future work including multi-actuator spatial rendering, cross-speed perceptual evaluation, and data-driven rendering.Source: §7 · pp. 14–15