Research Project

Deciphering human emotion communication through skin contact quantification

Subtle contact changes underlying different emotional percepts

March 18, 2022

Shan Xu Chang Xu, Sarah McIntyre, Håkan Olausson, Gregory J. Gerling

We routinely communicate distinct social and emotionalsentiments through nuanced touch. For example, we might gently hold another’sarm to offer a sense of calm, yet intensively hold another’s arm to expressexcitement or anxiety. As this example indicates, distinct sentiments may beshaped by the subtlety in one’s touch delivery. However, such contact detailsare still not fully understood due to the difficulty in quantifyingnaturalistic human touch.


This workinvestigates how slight distinctions in skin-to-skin contact influence both therecognition of cued emotional messages (e.g., anger, sympathy) and the ratingof emotional content (i.e., arousal, valence).


This work shows how touchers subtly, but significantly, vary the magnitudes of their skin-to-skin contact to convey distinct social messages. Besides improving receivers’ recognition of cued messages, this subtle tuning also correlates with receivers’ perception of underlying valence and arousal. For instance, arousal increases with velocity for stroking, and depth for holding. More interestingly, valence is tied with velocity, here for the case of human-to-human touch, which matches the trend that has been widely reported and reproduced in the case of brushing stimuli.

Experimentalsetup with 3D visual tracking system

(A) Touchers and receivers are separated by an opaque curtainand without verbal communication. Instructions and questions are displayed andrecorded by computers in front of participants. The depth camera (Kinect) wasset in between to capture hand-arm interactions. (B) A snapshot taken by the depthcamera during the experiments. The toucher was delivering a cued emotionalmessage to the receiver’s forearm. (C) An example 3D forearm point cloudand hand mesh. On the forearm, black line segments denote a normal vector perpoint, and red points illustrate the contact region. Six time-series contactattributes include: (D) contact area as the overall area of the contact regionon forearm, (E) indentation depth as the average indentation of all contactedhand points relative to the forearm surface, (F) absolute value of handvelocity and its three orthogonal components. The vector of spatial handvelocity is derived from the position of the middle metacarpophalangeal joint.Longitudinal velocity is along the direction from receiver’s elbow to wrist. Verticalvelocity follows the vertical direction pointing up. Lateral velocity isperpendicular to the other two velocities pointing left.


Delivered bythe same gesture, contact attributes change across emotional messages

(A) Contactattributes of messages delivered by touchers (top row) and recognized byreceivers (bottom row). One attribute that well distinguishes messages, i.e.,absolute velocity (cm/s) or indentation depth (cm), is displayed per gesture.(B) Importance of contact attributes derived by their contribution inclassifying successfully recognized messages. Vlg:longitudinal velocity (cm/s), Vvt: verticalvelocity (cm/s), Vlt: lateral velocity (cm/s), Vabs: absolutevelocity (cm/s), A: contact area (cm2), Dp: indentation depth(cm), Du: duration (s).

Absolutemeasurements per contact attribute, message, and gesture. Points denote trialdata.

Contact attributes encode valence andarousal states

(A) Valence and arousal ratings are similarbetween gestures per message. Blue cross marks represent valence and arousalratings retrieved from dataset using written word stimuli. (B) Contactattributes distinctly tune arousal and valence ratings within a gesture.Absolute velocity (cm/s) and indentation depth (cm) exhibit clear trends.




  • Xu, S., Xu, C., Mcintyre, S., Olausson, H., & Gerling, G. J. Subtle Contact Nuances in the Delivery of Human-to-Human Touch Distinguish Emotional Sentiment. [DOI] IEEE Transactions on Haptics. 2021
No items found.