John-John Cabibihan
Menu
On this pageBlind testSkinPress labPapers
ResearchOverview and atlas01Humanizing artificial touch02Social robots for autism03Gestures and social interaction04Prosthetic hands for war-wounded children05Tactile sensing for surgery and grasping06Robotic construction and architected materials07Smart materials, sensing and energy08Ride comfort and mobility ↗09Rehabilitation, VR and hydrotherapy
PatentsBookPublicationsCV
Chapter 01Pisa, Singapore and Doha

Humanizing artificial touch

A handshake or a pat on the arm says a lot, and most artificial hands say it badly. They are hard and cold, and people notice. This work asks what a synthetic skin needs, in softness, warmth and structure, before a touch from it passes for a human one.

588touches on the forearms of 28 blindfolded people
46%of touches from the warm artificial hand were taken for a human hand
28.4°Cthe skin temperature that 165 people preferred
18papers since 2006, and a 2026 book
Blind test

Was that a human hand?

Three hands touched the forearms of 28 blindfolded people with about 1 N of force: a real hand, a soft artificial hand at room temperature, and the same soft hand heated to 28.4°C. After each touch people rated how sure they were that it was human. The cold hand was easy to spot. The warm one could not be told apart from the real hand. Hover a segment for the counts.

Very sure prostheticSure prostheticNot sureSure humanVery sure human
half the answersHuman hand59% said humanWarm artificial, 28.4°C46% said human · AUC 0.56, p = 0.06Cold artificial, 21°C33% said human · AUC 0.70, p < 0.0001

Source: IEEE TNSRE, 2015, Table III. AUC compares each artificial hand with the human one; 0.5 means people could not tell them apart. There is a video of the test.

2015 · Qatar University, tested at NUS

A hand cast from a CT scan

The artificial hand was moulded from CT slices 0.5 mm apart, with a skeleton inside. A robot arm pressed 780 points on the real hand and on two replicas at 1 N each.

47%widest gap to the human hand, replica with bones; the narrowest was 14%
143%widest gap without bones; the narrowest was 54%
2015 · Qatar University, tested at NUS

Warm enough, for long enough

Heaters under the skin brought the surface to 28.4°C. A human fingertip in the same 21°C room measured 28.9°C on arrival and 27.8°C half an hour later, so the target sits inside the range a real hand drifts through.

It matters because people stop wearing prostheses they dislike. Rejection of hook prostheses has been reported as high as 50%.

Skin

Which sample feels like skin?

Before building the hand, 165 people touched a board of 16 silicone samples: four materials from soft to hard, each held at four temperatures. They chose the one that felt most like human skin. Choices piled up on the softest material at 27°C and 32°C. Hover or tap a sample.

AShore 30BShore 55CShore 70DShore 86
22°C
27°C
32°C
37°C

Material A at 27°C

Ecoflex OO-30, Shore OO 30. The softest of the four.

36 of 165 people picked this sample

Picking at random gives each sample a 1 in 16 chance, about 10 people. The three marked cells drew 90 of the 165.

Press lab

A finger that gives way

In a handshake, the middle phalanges of the little, ring and middle fingers take most of the force. Pressed at 2 N, these human phalanges dent by more than 3 mm. A commercial prosthetic finger dents by 0.2 mm. The 2011 study at NUS simulated synthetic phalanges with and without a hollow pocket under the pad. Pick a material and a core, and watch the indenter press.

Drag to turn. The cut face shows the section.

Skin material

Inside the finger

A solid core, the way most cosmetic fingers are made.

Dent at 2 N 0.77 mm

0 mm1 mm2 mm3 mm4 mmThis designHuman finger, 10 peopleCommercial prosthetic0.773.3 to 3.80.2

Only the 2 mm pocket gets past 2 mm, close to what a human finger gives under a light touch.

Source: Journal of NeuroEngineering and Rehabilitation, 2011. Human values are means for 10 people, flat hand. The model shows the simulated depth; the dent's shape is drawn for illustration.

Under the skin

Fingertips, ridges and thickness

2008 · NUS

Measuring the human fingertip

Silicone and polyurethane fingertips were compared with human ones for compliance, conformance and hysteresis, in experiments and finite element models. None of the synthetic skins matched the human fingertip.

International Journal of Social Robotics

2011 · NUS

Ridges, like fingerprints

A ridged skin over a 2×2 tactile array helped tell three surface curvatures apart. Of the three classifiers tried, the best reached 97.5% accuracy.

Sensors

2014 · QU

How thick a skin can be

Flat, curved and Braille surfaces were pressed through 1, 3 and 5 mm of skin, simulated and checked with pressure film. Under 1 and 3 mm the shapes stayed distinct. Under 5 mm they blurred together.

IEEE Sensors Journal

2026 · Book

The book

The skins, sensors and structures from this line of work are collected in a book from Elsevier. Every chapter has a working figure on its own page.

Explore the 17 chapters · Elsevier

Record

Papers and grants

18 papers

Grants

  • 2015Neuromorphic Tactile Sensing: A Paradigm Shift for Prosthetics and Robotic Surgery
    US$823,782 · Qatar National Research Fund · Lead Principal Investigator
  • 2010Affective Tele-Touch Technology
    US$111,085 · National University of Singapore · Lead Principal Investigator
  • 2008Design of Prosthetic Skins with Humanlike Softness
    US$141,342 · National University of Singapore · Sole Investigator