Perceived Realism of Virtual Textures Rendered by a Vibrotactile Wearable Ring Display

Abstract

Wearable haptic displays that relocate feedback away from the fingertip provide a much-needed sense of touch to interactions in virtual reality, while also leaving the fingertip free from occlusion for augmented reality tasks. However, the impact of relocation on perceptual sensitivity to dynamic changes in actuation during active movement remains unclear. In this work, we investigate the perceived realism of virtual textures rendered via vibrations relocated to the base of the index finger and compare three different methods of modulating vibrations with active finger speed. For the first two methods, changing finger speed induced proportional changes in either frequency or amplitude of vibration, and for the third method did not modulate vibration. In psychophysical experiments, participants compared different types of modulation to each other, as well as to real 3D-printed textured surfaces. Results suggest that frequency modulation results in more realistic sensations for coarser textures, whereas participants were less discerning of modulation type for finer textures. Additionally, we presented virtual textures either fully virtually in midair or under augmented reality in which the finger contacted a flat surface; while we found no difference in experimental performance, participants were divided by a strong preference for either the contact or non-contact condition.

Full text

1
Perceived realism of virtual textures rendered by a
vibrotactile wearable ring display
Rebecca Fenton Friesen*, Member, IEEE, and Yasemin Vardar, Member, IEEE
Abstract—Wearable haptic displays that relocate feedback
away from the fingertip provide a much-needed sense of touch to
interactions in virtual reality, while also leaving the fingertip free
from occlusion for augmented reality tasks. However, the impact
of relocation on perceptual sensitivity to dynamic changes in
actuation during active movement remains unclear. In this work,
we investigate the perceived realism of virtual textures rendered
via vibrations relocated to the base of the index finger and com-
pare three different methods of modulating vibrations with active
finger speed. For the first two methods, changing finger speed
induced proportional changes in either frequency or amplitude of
vibration, and for the third method did not modulate vibration.
In psychophysical experiments, participants compared different
types of modulation to each other, as well as to real 3D-printed
textured surfaces. Results suggest that frequency modulation
results in more realistic sensations for coarser textures, whereas
participants were less discerning of modulation type for finer
textures. Additionally, we presented virtual textures either fully
virtually in midair or under augmented reality in which the
finger contacted a flat surface; while we found no difference in
experimental performance, participants were divided by a strong
preference for either the contact or non-contact condition.
Index Terms—Surface Haptics, Texture, Wearables, Virtual
Reality
I. INTRODUCTION
RECENT advances in virtual reality interfaces allow us
to explore immersive virtual worlds and complex ob-
jects through rich visual and auditory feedback, yet haptic
interactions with virtual stimuli remain far more primitive.
Commercially available haptic displays attempting to close
this gap, such as vibrating handheld controllers [1], [2] or
gloves [3], [4], generally provide haptic feedback directly
to the fingertips or palm. While these locations are where
a user would expect sensations during active touch, such
placements can be particularly problematic for augmented
and mixed reality: occluding the fingertip with gloves or
other hardware demonstrably reduces tactile acuity [5], [6],
and bulky hardware concentrated in the small workspace of
the hand can hamper dexterity. One approach to mitigate
these deleterious effects is to ”fold away” actuators when not
needed, and apply them to the fingertip only during virtual
interactions [7]. An alternative solution is the use of relocated
feedback, in which haptic stimulation that would normally
occur at the fingertip or hand is permanently relocated to a
Rebecca Fenton Friesen is with the Department of Mechanical Engineering,
Texas A&M University, College Station, Texas and the Department of
Cognitive Robotics, Delft University of Technology, Delft, The Netherlands
Email: rfriesen@tamu.edu
Yasemin Vardar is with the Department of Cognitive Robotics, Delft
University of Technology, Delft, The Netherlands. Email: y.vardar@tudelft.nl
* Corresponding author.
Manuscript received Month October, 2022
more convenient location for actuation such as the wrist or
proximal parts of the hand, which provides a larger work area
and leaves the fingertip free for additional tasks.
Relocated feedback has proven promising for several types
of haptic interactions [8], particularly when using squeeze or
shear forces as intelligible cues of contact and softness of
virtual objects [9]–[11]. Of particular interest to us is the
relocation of texture-induced vibrations, traditionally applied
via a grasped stylus [12]–[14] or directly to the fingertip [15]–
[17] to mimic the sensation of interacting with a textured
surface. While the aforementioned works repeatedly demon-
strate that people can identify distinct vibration patterns as
different textures and are sensitive to changes in frequency,
intensity, and spectral complexity, preliminary research [18] is
still exploring whether such rich frequency information even
remains intelligible when relocated away from the fingertip.
Other research groups are focusing primarily on device design
of low-profile wearable rings, demonstrating compact methods
of actuation, broadband vibratory feedback, and dynamic
stimulation for both navigation cues [19] and texture display
[20], [21].
A significant challenge when designing small vibrating
wearable devices for realistic texture rendering is accounting
for the large changes in the frequency content of real texture-
induced vibrations as the active finger dynamically moves and
changes speed [22]. In order to mimic these real interactions,
vibration frequencies must modulate with fingertip speed such
that spatial frequencies remain constant [23]. This rendering
method often relies on measured position of the finger to pre-
serve spatial constancy of texture patterns [24], with changes
to temporal frequency a direct result of changes in speed.
However, such implementation results in several practical
challenges; firstly, one must account for the strong resonances
of many vibrotactile actuators that result in dramatic changes
in intensity that couple with any change in frequency [25].
Additionally, preserving spatial frequency requires monitoring
finger speed and updating actuator output at a rate much
faster than would be necessary for alternative methods such as
amplitude envelope modulation. We hypothesize that for some
texture-mimicking vibrations, particularly when relocated to
less sensitive areas away from the fingertip, the additional
challenges of preserving spatial frequency through continuous
variation of temporal frequency may be unnecessary. We ask
whether modulating amplitude instead, which still preserves
expected changes in spectral power [23], could result in
equivalently realistic sensations of texture.
In this study, we explore the impact of several particular
velocity-dependent signal modulation schemes on perceived
realism and discrimination of virtual textures for a relocated

2
vibrotactile display worn as a ring. Three different modulation
types were considered; (1) varying the vibration frequency as
finger velocity changed, analogous to maintaining a constant
spatial frequency, (2) varying the amplitude of vibration with
finger velocity, such that signal power proportionally increases
with increasing velocity, and (3) no modulation of vibration at
all while the finger is moving, regardless of scanning velocity.
We tested these modulation schemes with both fine and coarser
(i.e. high and lower frequency) virtual textures, as well as
when the finger was or was not receiving additional surface
cues via contact with a flat surface. We looked for impacts on
judgments of realism and pleasantness of the virtual textures,
both in comparison to real textures and compared to each other.
II. SY ST EM DESIGN
The following section details our construction and control
of virtual textures applied via a wearable vibrotactile ring. We
first present the design and characterization of the wearable
ring, as well as the wider experimental apparatus for tracking
participant movement and controlling the ring output. We then
describe the virtual textures and types of velocity modulation
used in this study, and the real textures used for comparison. In
the subsequent section, these virtual stimuli and real textures
were compared and the impact of modulation type, ampli-
tude, and frequency of virtual textures on perceived realism
and pleasantness was assessed in a series of psychophysical
experiments.
A. Wearable Ring
Our primary design considerations for the wearable texture
ring display were the location, size, and frequency response
of the actuator. We chose to locate actuation on the dorsal
side of the proximal phalanx of the index finger, as shown in
Fig. 1a, leaving the fingertips and grasp area (i.e. ventral side
of the hand) relatively unencumbered yet otherwise keeping
actuation close to the fingertip. The actuator used in this study
was the HapCoil-One (Tactile Labs), measuring 11x11x37 mm
and attached so as to vibrate in the direction along the finger,
laterally against the skin. We chose this actuator as it has a
stated bandwidth of 10-1000 Hz, spanning the majority of the
human tactile perceptual range (0-1000 Hz) [26]. Additionally,
it is small enough to fit within the approximately 15x40 mm
workspace of an index finger’s proximal phalanx.
We characterized the actuator behavior using a single-
point laser doppler vibrometer (Polytec OFV-5000, OFV-505
sensor head). All measurements aligned with the direction of
vibration and were taken while the actuator was in contact
with the hand (see Fig. 1a). The actuator was secured to the
index finger using a Velcro strap with an integrated force
sensor (FSR 402, Interlink Electronics) and tightened to apply
a 0.5 N squeezing force. The input voltage of the actuator was
amplified with a class D audio amplifier, the AudioAmp 2
Click (Mikro Electronika) with 20 dB gain. The HapCoil-
One has a reported resonant frequency of 65 Hz, observed in
the peak actuator displacement plotted in Fig. 1b. In order
to achieve standardized actuator displacement regardless of
commanded frequency, we passed our output voltage through
a 400th order zero phase arbitrary response filter, hand-tuned
in Matlab 2021 (see Fig. 1c).
The addition of the filter considerably flattened the displace-
ment magnitude; the steep roll-off in amplitude below 30 Hz
is due to the high pass filtering of the amplifier, necessary to
protect the actuator from DC voltage offsets. Performance of
the filter was observed at three different base voltage levels,
as shown in Fib. 1b, to demonstrate linearity across multiple
amplitudes.
50 100 150 200 250 300
frequency (Hz)
peak actuator displacement (mm)
unfiltered
filtered
(a) (c)
1.5 V
1 V
0.75 V
10
-1
10
-2
direction of
vibration measured
displacement
(b)
filter magnitude
frequency (Hz)
10
-1
10
-2
50 100 200 300
250
150
Fig. 1. (a) Placement of the vibrotactile actuator and measurement location
of the generated vibrations via laser doppler vibrometer. (b Peak actuator
displacement as a function of input frequency for three different input
amplitudes. (c) Filter for flattening actuator resonant peak.
Velocity-dependent haptic rendering requires real-time ve-
locity estimation. In this experiment, we tracked finger po-
sition, and therefore velocity, with a one-degree-of-freedom
pulley system shown in Fig. 2a-c. The bottom of the wearable
ring clipped into a magnetic bracket on the thin nonelastic
fishing line stretched between two pulleys and a quadrature
encoder shaft. As a user moved their finger back and forth
across the texture samples, the encoder provided a resolution
of 10.6 µm, monitored at 10 kHz.
B. Real Textures
We designed and 3D printed a set of textured surfaces for
use in comparison tests with our virtual texture display. The
real textures consisted of 3D printed resin using stereolithog-
raphy (manufactured by 3Delft), and all had a 20 mm x 50 mm
surface area. We deemed the 50 mm length long enough for
a user to swipe their finger along, but short enough to fit two
textures side by side for comparison in our display.
All textures consisted of a single spatial frequency compo-
nent that varied along the longest axis. We originally designed
textures with a spatial frequency as high as 2 mm−1, or 0.5 mm

3
160
80
0
0
45
speed
(mm/s)
desired actuator displacement voltage to actuator
time (s)
0 0.5
modulation types
(1)
(2)
(3)
finger tracking
position
(mm)
filter
Hz
magnitude
time (s)
0 0.5
virtual
real
encoder
(a) (b)
(c)
encoder
input
(position)
ltered
derivative
(velocity) absolute
value
(speed)
frequency-,
amplitude-, or
no-modulation
400th
order
lter
voltage
output
to ring
Fig. 2. (a) The experimental setup with a participant feeling both a real and virtual texture side by side. (b) Simplified diagram of Simulink Realtime model,
measuring encoder input and generating voltage output at 10kHz. (c) Pictorial progression of virtual texture rendering. 10kHz sampled encoder output and its
filtered derivative provide position and speed of the finger. Paired with equations 1-3, this generates a virtual texture using one of three different modulation
types: 1- frequency modulation (FM), 2- amplitude modulation (AM), and 3- no modulation (NM). The desired displacement signal is passed through a 400th
order zero phase arbitrary response filter to remove amplification effects of actuator resonance. Finally, the filtered voltage signal is sent to the actuator to
produce a vibration with the desired displacement.
ridge spacing. However, this spacing proved so fine that it
elicited no noticeable vibration on a sliding finger, instead
serving only to reduce the overall friction coefficient. We
therefore selected real textures coarser than this value, while
still fine enough to induce vibrations within the bandwidth of
our vibrotactile actuator. The final set consisted of 0.5mm−1,
1 mm−1, and 1.5 mm−1spatial sinusoids; see Fig. 3 for close-
up photos of each texture. All textures had a peak-to-trough
height of 1 mm.
In order to ascertain that these textures do elicit vibrations
that correspond with their spatial frequency, we measured the
lateral force between a finger and the textured surface during
active scanning. Textures were mounted on a 6-axis force
sensor (ATI Nano17 Titanium), and the first author swiped
each sample for 5 seconds with a normal force trained to
average 0.4 N and a scanning speed averaging 80 mm/s. Speed
was regulated using a metronome, so in practice varied widely
as the finger changed direction back and forth across the
texture sample. Despite non-precise instantaneous force and
speed, all three textures induce the expected vibrations, as
can be seen in both the time domain snapshot and summary
spectral data in Fig. 3.
C. Virtual Textures
In the context of this study, all virtual textures are vibrations
applied to the proximal phalanx of the right index finger
during movement, consisting of a single frequency compo-
nent corresponding to each of the real texture samples. We
100 200
0
1
2
100 200
0
1
100 200
0
1
40 Hz
80 Hz
120 Hz
frequency (Hz)
amplitude (mN)
-50
0
50
0 0.1
-50
0
50
-50
0
50
0 0.1
0 0.1
time (s)
30
30
30
texture #1
texture #2
texture #3
amplitude (mN)
1.5/mm
1.0/mm
0.5/mm
Fig. 3. Cropped photos of three real texture samples, all with different spatial
frequencies but the same 1 mm height. To the left are shown the corresponding
time and frequency domain responses of the lateral force between a finger and
the texture measured while the finger moves over each surface with an average
speed of 80 mm/s and normal force of 0.4N. The expected temporal frequency
corresponding to the spatial frequency scanned at 80 mm/s is indicated in red.

4
tested three types of velocity-dependent modulation of virtual
textures S1−3:
S1(x,t) = Asin(2πfsx) = Asin(2πf(˙x)t)(1)
S2(x,t) = A(˙x)sin(2πf@80t)(2)
S3(x,t) = Asin(2πf@80t)|˙x|>0,
0 ˙x=0.(3)
Here, A,t,x,fs,frefer to vibration signal amplitude, time,
finger displacement in lateral direction, and spatial and tem-
poral frequency of a texture, respectively. f@80 represents the
time-domain frequency value having the maximum amplitude
of the vibration that occurred while a finger scans a particular
real texture with a speed of 80 mm/s.
In the first type, shown in Equation 1, the vibration is
spatially defined. In other words, the temporal frequency of
the vibration ( f) varies with finger velocity (˙x) such that
their multiplication is equal to the spatial frequency ( fs) of
the corresponding surface. This modulation type most closely
matches the frequency change expected while scanning a
real texture with spatially distributed surface features. This
modulation method is named frequency-modulation (FM) in
the rest of the manuscript.
The second type of modulation, shown in Equation 2, also
modulates vibratory power with scanning velocity, but in a
different way. Here, we vary only the intensity of vibration
with scanning velocity, while the temporal frequency ( f@80 )
remains unchanged at the peak component of the correspond-
ing texture measured at 80 mm/s (check Fig. 3 for the values).
In implementation, we proportionally increase the amplitude
(A) linearly with increasing scanning speed from 0 to 80 mm/s;
at faster scanning speeds, amplitude saturates at the same
value used for the other modulation types. The particular
speed threshold value at which amplitude no longer changes
was somewhat arbitrary but was inspired by the plateauing
of measured vibratory power at high speeds for real texture
interactions [23]. We hereafter refer to this modulation type
as amplitude-modulation (AM).
In contrast, the method in Equation 3 demonstrates almost
no speed dependence at all; a moving finger feels a sinusoidal
vibration at the peak temporal frequency of the corresponding
texture measured at 80 mm/s (Fig. 3) that does not change
in frequency or amplitude, aside from turning off when the
finger is completely still. This method is called no modulation
(NM). All three types of modulation are graphically depicted
in Fig. 2c.
When choosing vibration amplitude, we sought to minimize
intensity differences between the real textures and their vir-
tual analogs to avoid overwhelming potentially more subtle
differences in modulation type. Matching perceived intensities
of texture-induced vibrations on the fingertip to vibrotactile
vibrations applied on the base of the finger is non-trivial,
especially across a large participant population, and outside
the scope of this project. Instead, we chose to modulate virtual
texture amplitudes between different frequencies to roughly
the same ratios as seen in real texture force measurements in
Fig. 3. Just as increasing the spatial frequency of a real texture
reduces the peak forces on the fingertip (compare peak values
in Fig. 3), increasing the frequency of our virtual textures
reduces the driving voltage and therefore peak displacement
of the actuator. For all real-to-virtual comparisons in this
work, this ratio corresponds to 1, 1.5, and 2 V peak values
for the high, mid, and low-frequency vibrations. Absolute
maximum and minimum voltages were determined by the
study authors, chosen to be perceptible yet not uncomfortably
strong. Participants’ amplitude preferences for each virtual
texture frequency are further explored in the virtual-to-virtual
texture comparisons.
We implemented velocity-dependent modulation of virtual
textures using a NI PCIe-6323 DAQ and the Simulink Desk-
top Real-time environment. Finger position was sampled at
10 kHz, passed through a discrete derivative block and filtered
(TC = 0.01), and absolute valued to find finger speed. Position
or speed determined actuator output as defined in Equations 1-
2. Finally, the input voltage of the actuator was filtered to
obtain the desired displacement by accounting for actuator
resonance. This procedure is summarized in Fig. 2b-c.
III. PSYCHOPHYSICAL EXPERIMENTS
20 people (two left-handed, two women, ages 22-36) par-
ticipated in this study. This study was approved by the Ethics
Board of Delft University of Technology with case number
1781. All participants were students or employees of the
university.
A. Training Procedure
Prior to each participant session, all experimental surfaces
and the wearable ring were disinfected in accordance with
approved coronavirus safety procedures. All participants were
fitted with the wearable ring on their right index finger
with a squeezing force of 0.5 N, and wore noise-canceling
headphones playing continuous pink noise to mask potential
sound effects from virtual texture actuation. The headphones
also provided audio cues signaling the trial start and end
times, as well as a metronomic beep during trials to ensure a
consistent scanning speed across participants.
Applied Force: In order to reduce the effects of widely vari-
able normal force, all participants practiced scanning a sample
real texture for one minute while applying a 0.4 N pressing
force to the surface. For this training, the sample texture was
mounted on the force sensor (ATI Nano17 Titanium). During
practice, they were provided with a visual graph of their real-
time applied force.
Scanning Velocity: Subsequent to practicing controlling the
contact force on real textures, participants practiced mov-
ing at a prescribed average scanning velocity for both real
and virtual textures. The wearable ring was attached to the
magnetic clip on the position tracking line, and participants
practiced scanning real and virtual textures, both 50 mm in
length while timing each 50 mm long swipe to a metronome
beat played over the headphones. The metronome ensured a
similar average scan velocity, in order to keep the temporal

5
frequencies caused by real and spatially determined virtual
textures similar across all participants. Following feedback
from a pilot study, we chose an average swipe speed of
80 mm/s, which when paired with the 50 mm texture length
corresponds to a metronome speed of 96 bpm.
Free Magnitude Estimation: Finally, participants were
guided through five practice trials comparing the coarsest
real texture to virtual textures of various frequencies and
modulation types, in order to become familiar with the sim-
ilarity rating system. Each practice trial began with an ”on”
tone, then 30 seconds of metronome beats during which the
participant could feel either texture, as long as their swipes
were aligned with one texture length and a metronome period.
At the end of the trial, an ”off” tone played, and participants
were instructed to provide a free magnitude estimation of
the similarity between the pair. The rating for the first trial
was arbitrary; subsequent trials that were less similar would
be rated lower, and trials perceived as twice as similar as a
previous pair should be rated twice as high. Ratings could
be comprised of any non-zero positive number, including
fractions. Participants could practice generating free magni-
tude estimations of similarity for more than 5 trials if they
wished until both they and the experimenter agreed that they
understood the concept.
B. Experiment 1: Virtual and Real Textures
The goal of Experiment 1 was to compare the similarity
between real and virtual textures for different virtual modula-
tion types, across different texture length scales, and contact
conditions. In order to accommodate the large number of
variables, we split this experiment into two 15-trial sets,
performed before and after Experiment 2. The two sets differed
only in contact conditions; for one set, participants felt all
virtual textures while in contact with a 50 mm long flat surface
printed in the same resin material as the real textured surfaces.
For the other set, participants felt all virtual textures under
non-contact conditions, in which their finger swiped over a
50 mm rectangular hole in the presentation plate (see Fig. 2a).
Half of the participants performed the first set in contact
and the second in non-contact, while for the other half the
contact conditions were reversed. During the block of 15
trials, participants were instructed to consider the similarities
between pairs of previous trials when evaluating the similarity
between real and virtual textures in each subsequent trial.
This approach ensured that participants’ judgments remained
consistent throughout each trial.
The 15 trials within each set were divided across the three
real textures. The first five trials, presented in randomized
order, compared the coarsest 0.5 mm−1real texture to all
three low-frequency virtual textures using each of the three
modulation types, as well as all three frequency-modulated
(FM) virtual textures. Note that the combination of the three
modulation types for one frequency and three frequencies
of one modulation type results in five total trials due to
overlap between the two sets. The next five randomized
trials consisted of the same type of comparisons, but with
the real texture replaced with the 1 mm−1sample, and the
different modulation types replaced with the corresponding
mid-frequency virtual textures. Similarly, the final five trials
compared the finest 1.5 mm−1real texture to the corresponding
high-frequency virtual textures of all modulation types, along
with all frequencies of FM virtual textures. Table I summarizes
all the comparisons made in Experiment 1.
TABLE I
EXP ERI ME NT 1CO MPAR ISO NS
Trial # Real texture Virtual texture comparisons
1-3 low spatial freq. low frequency (all modulation types)
4-5 low spatial freq. medium and high frequency (only
FM)
6-8 medium spatial freq. medium frequency (all modulation
types)
9-10 medium spatial freq. low and high frequency (only FM)
11-13 high spatial freq. high frequency (all modulation types)
14-15 high spatial freq. low and medium frequency (only FM)
C. Experiment 2: Virtual Texture Comparison
For Experiment 2, participants once again made free magni-
tude estimations of similarity for texture pairs, but all textures
were virtual and presented under non-contact conditions. In
addition to rating similarity, participants were asked to indicate
which texture was more ”real”, i.e. more similar to a real
textured surface, and which was more ”pleasant” to touch. In
asking for judgments of pleasantness, we sought an additional
qualitative measure of the virtual texture experience; in partic-
ular, we were curious if pleasantness mirrored judgments of
realism. If participants perceived no realness or pleasantness
differences, they were asked to choose a texture at random.
This experiment also consisted of 15 comparison trials, and
both the trial order and left/right placement of each texture
were randomized. Throughout the set, participants compared
different modulation types to each other, different amplitudes
to each other, and different frequencies to each other; see
Table II and Section IV-D for details of the selected stimuli
set.
TABLE II
EXP ERI ME NT 2COMPARISONS
Trial # Pairwise
Comparisons
Constant Parameters
1-3 Modulation
type
Amplitude = medium, Spatial frequency =
low
4-6 Modulation
type
Amplitude = medium, Spatial frequency =
high
7-9 Amplitude Modulation = NM, Spatial frequency = low
10-12 Amplitude Modulation = NM, Spatial frequency = high
13-15 Frequency Amplitude = medium, Modulation = NM
D. Post-experiment Questionnaire
Following Experiments 1, 2, and a repeat of Experiment
1 under the alternate contact condition, participants removed
the wearable ring and answered a short survey. They were
asked: ”What made a virtual texture feel more or less like
a real textured surface?” and ”What made a virtual texture
feel more or less pleasant?”, as well as what effect contact or

6
non-contact conditions for experiment 1 had on virtual texture
perception.
IV. RES ULT S
A. Scanning Speed
Standardizing scanning speed across participants was impor-
tant to ensure uniformity of comparisons between a spatially
determined real or virtual texture and a virtual texture in which
the temporal frequency did not change. Fig. 4 summarizes
the actual scanning speeds of participants throughout the
experiments and confirms that average speeds were close to
the desired 80 mm/s imposed by the metronome. Raw data
from several participants in the first experiment highlights that
actual velocities predictably varied quite widely during touch,
as participants had to slow, stop, and turn around at the end
of each swipe.
0 0.5 1 1.5
0
50
100
150
200 5th participant
10th
15th
20th
time (s)
finger speed (mm/s)
experiment #1
experiment #2
experiment #1
(2
nd
round)
Fig. 4. Summary of participant scan speeds during all experimental rounds.
The targeted speed of 80 mm/s is indicated with the dotted line. On the left,
raw speed measurements are shown for four representative participants over
several swipes. A summary of average speeds across all participants for each
experiment is shown on the right.
B. Participant Ratings
Since each participant created their own range of similarity
ratings, we normalized similarity ratings across all partici-
pants for each 15-trial experimental block using geometric
normalization procedure [27]. Accordingly, each participant’s
response was normalized by dividing by the participant’s mean
in the given experiment, then multiplying by the grand mean
for all participants. Ratings were not normalized across the
experimental blocks since each block consisted of substantially
different conditions: virtual textures compared to either real
stimuli or other virtual textures.
C. Virtual Textures Compared to Real Textures
We first investigated whether equivalent spatial frequencies
resulted in higher similarity ratings between real and virtual
texture pairs. We compared normalized similarity ratings for
the subset of trials in which each spatially determined virtual
texture (i.e., a frequency-modulated texture) was compared to
each real texture. We first made a Shapiro-Wilk test and con-
firmed that all distributions passed the normality test. Then, we
conducted a three-way ANOVA with repeated measures to test
the effects of contact condition, real texture spatial frequency,
and virtual texture spatial frequency on the similarity ratings.
The results showed that contact conditions did not sig-
nificantly affect the perceived similarity of real and virtual
textures (F(1,19) = 0.002, p=0.96). However, both real
(F(2,38) = 7.8, p=0.001)and virtual (F(2,38) = 18.51,
p<0.001)texture spatial frequencies and their interaction
(F(4,76) = 32.53, p<0.001)significantly affected the simi-
larity ratings. Fig. 5 shows the confusion matrix between real
and virtual textures for the three spatial frequencies tested. The
values represent the mean normalized similarity ratings across
all participants and contact conditions. In general, participants
rated equivalent spatial frequencies as more similar, and stim-
uli pairs were rated increasingly dissimilar as the frequencies
increasingly differed. The highest frequency virtual texture
was an exception to this trend, as it was consistently rated
as more similar to the 1 mm−1real sample.
8.6
7.8
6.3
6.5
8.7
9.7
5.0
6.7
9.0
real texture spatial frequency
virtual texture
spatial frequency
0.5 mm-1 1 mm-1 1.5 mm-1
0.5 mm-1
1 mm-1
1.5 mm-1
more
similar
less
similar
Fig. 5. Confusion matrix of rated similarities between real textures and
frequency modulated virtual textures. Spatial frequencies of the virtual and
real textures are indicated along the axes, and the normalized similarity ratings
averaged across all participants and contact conditions are listed along with
their corresponding levels of shading.
Next, we investigated the impact of the modulation method
on rated similarity, both for differing length scales and contact
conditions. Similarity ratings between virtual and real texture
pairs with equivalent spatial frequencies are summarized in
Fig. 6; for AM and NM conditions, the assumed spatial
frequency is the temporal frequency divided by the average
scan speed of 80 mm/s. First, we confirmed that almost all dis-
tributions passed the normality assumption via a Shapiro-Wilk
test. Then, using a three-way ANOVA with repeated measures,
we assessed the significance of similarity differences caused
by modulation type, spatial frequency, and contact condition.
The results revealed that contact conditions did not sig-
nificantly affect the perceived similarities between real and
virtual textures (F(1,19 =0.064, p=0.804). Nonetheless,
the effects of modulation type (F(2,38) = 8.16, p=0.001),
spatial frequency (F(2,38) = 6.03, p=0.005), and their
interaction (F(4,76) = 14.13, p=0.007)in the perceived
similarities were statistically significant. Bonferroni corrected
post hoc paired t-test showed that modulation type only had
a significant impact on the similarity ratings at the lowest
frequency; statistically different pairs (p<0.05)are marked

7
in Fig. 6.
0
5
10
15
similarity ratings
non-contact contact
0.5 mm-1 1 mm-1 1.5 mm-1
0
5
10
15
similarity ratings
*
*
*
FM AM NM
Fig. 6. Normalized similarity ratings between virtual and real texture
pairs with equivalent spatial frequencies. The results corresponding to each
experimental condition are color-coded. Individual participant responses and
the mean are shown in gray and black dots, respectively, while box plots
summarize the distribution. The central lines show the medians; box limits
indicate the 25th and 75th percentiles. The whiskers extend to 1.5 times the
interquartile range. Statistically significant pairwise comparisons (p<0.05)
are indicated with asterisks, ∗.
D. Virtual Texture Comparisons
For the second set of experiments, participants compared
different virtual textures to each other under non-contact con-
ditions. Of these 15 trials, a subset was designed to compare
modulation types, another to compare different amplitudes,
and one to compare different frequencies. We first conducted
Shapiro-Wilk tests and confirmed that almost all distributions
passed the normality test. Then, we analyzed each subset
separately using ANOVA with repeated measures and post hoc
Bonferroni-corrected paired t-tests; the results are summarized
below:
1) Modulation Comparisons: For six trials, study partic-
ipants compared modulation types for both low frequency
(0.5 mm−1) and high frequency (1.5 mm−1) virtual textures
at the same amplitude (1.5 V). Normalized similarity ratings,
realness, and pleasantness judgments are summarized in Fig.
7.
The results of two-way ANOVA with repeated measures
revealed that compared modulation types (F(2,38) = 9.30,
p<0.001)and texture frequency (F(1,19) = 7.39, p=0.014)
significantly affected the perceived similarities, but there was
no interaction between them (F(2,38) = 3.25, p=0.05). The
significantly different pairs (p<0.05 or p<0.01)are depicted
in Fig. 7. These differences are also reflected in trends across
the two frequency conditions: for low-frequency texture pairs,
AM and NM conditions were more similar to each other, while
FM conditions were rated less similar to both. In contrast, for
high-frequency textures, participants rated NM as similar to
both FM and AM, but FM and AM were less similar to each
other.
Pie charts indicate which percentage of participants selected
one or the other texture as more real and pleasant; most
strikingly, modulation of any type usually results in more
realism and pleasantness than no modulation.
similarity ratings
Modulation Type Comparisons
more
real
more
pleasant
0.5 mm-1 1.5 mm-1
**
* * * *
0
5
10
15
FM AM NM
Fig. 7. Similarity ratings, realism, and pleasantness choices for pairs of
virtual stimuli differed in modulation type (FM-AM, FM-NM, or AM-NM).
All amplitudes were the medium value. The tested virtual texture spatial
frequencies correspond to low or high values (0.5 mm−1 or 1.5 mm−1).
The results corresponding to each experimental condition are color-coded.
Individual participant responses and the mean are shown in gray and black
dots, respectively, while box plots summarize the distribution. The central
lines show the medians; box limits indicate the 25th and 75th percentiles. The
whiskers extend to 1.5 times the interquartile range. Statistically significant
pairwise comparisons are marked with asterisks; ∗and ∗∗ mean p<0.05 and
p<0.01, respectively.
2) Amplitude Comparisons: Virtual textures (0.5mm−1and
1.5 mm−1) with different amplitudes were also compared,
keeping the modulation type the same (NM); see Fig. 8 for
the summary of the results.
The results of two-way ANOVA with repeated measures
showed that amplitude (F(2,38) = 22.91, p<0.001)and
texture frequency (F(1,19) = 10.55, p=0.004)significantly
affected the perceived similarities, but there was no interaction
between them (F(2,38) = 0.114, p=0.89). The statistically
significant pairwise comparisons (p<0.05 or p<0.01)are
indicated in Fig. 8. Unsurprisingly, the greatest differences in
amplitude result in the lowest similarity ratings.
For low-frequency virtual textures, the two higher ampli-
tudes tended to be chosen as more realistic, with the medium
amplitude found most pleasant. For high-frequency textures,
the lowest amplitude was both more realistic and pleasant.
3) Frequency Comparisons: Only three trials compared
textures of different frequencies, all at medium amplitude and
NM; see Fig. 9 for the results.
A one-way ANOVA with repeated measures revealed that
texture frequency significantly affected the perceived similar-
ities (F(2,38) = 32.63, p<0.001). Statistically significant
(p<0.001)pairwise comparisons are marked in Fig 9.
Similarly to amplitude comparisons, greater differences in
frequency resulted in lower similarity ratings. Moreover, par-
ticipants appeared to find the middle frequency more realistic

8
Amplitude Comparisons
0.5 mm-1 1.5 mm-1
similarity ratings
more
real
more
pleasant
*
* *
*
* *
*
0
5
10
15
low med. high amplitude
Fig. 8. Similarity ratings, realism, and pleasantness choices for pairs of virtual
stimuli differed in amplitude (low-medium, low-high, or medium-high). No
modulation was applied for these pairs. The tested virtual texture spatial
frequencies correspond to low or high values (0.5 mm−1 or 1.5 mm−1).
The results corresponding to each experimental condition are color-coded.
Individual participant responses and the mean are shown in gray and black
dots, respectively, while box plots summarize the distribution. The central
lines show the medians; box limits indicate the 25th and 75th percentiles. The
whiskers extend to 1.5 times the interquartile range. Statistically significant
pairwise comparisons are marked with asterisks; ∗and ∗∗ mean p<0.05 and
p<0.01, respectively.
than both higher and lower alternatives, and overwhelmingly
more pleasant than the highest frequency.
Frequency Comparisons
similarity ratings
more
real
more
pleasant
0.5 mm-1
1 mm-1
1.5 mm-1
0
5
10
15
*
*
*
*
Fig. 9. Similarity ratings, realism and pleasantness choices for pairs of virtual
stimuli differed in frequency (0.5-1 mm−1, 0.5-1.5 mm−1, or 1-1.5 mm−1).
All amplitudes were at the medium value, and no modulation was applied.
The results corresponding to each experimental condition are color-coded.
Individual participant responses and the mean are shown in gray and black
dots, respectively, while box plots summarize the distribution. The central
lines show the medians; box limits indicate the 25th and 75th percentiles. The
whiskers extend to 1.5 times the interquartile range. Statistically significant
(p<0.01) pairwise comparisons are marked with asterisks ∗∗.
E. Participant Feedback
Following the three experimental rounds, participants had a
range of opinions on what improved the realism and pleasant-
ness of virtual textures, but some common themes emerged:
Six participants thought that the non-contact condition for
virtual texture resulted in more realistic virtual texture ren-
dering, some quite strongly, while 10 thought contact with a
flat surface improved realism (the remaining were unsure or
answered ”depends”). A common reason for preferring non-
contact was that the sensation of flat surface contact ”clashed”
with the actuation provided at the base of the finger, while
those who preferred contact stated that the additional sensation
on the fingertip helped. As over half the participants (12)
commented on the effect of contact on realism unprompted
early in the survey, this warrants further investigation.
V. DISCUSSION
In this study, we investigated the effect of signal modulation
methodology on the perceived realism of virtual textures
rendered via a vibrotactile display stimulating the proximal
phalanx of the index finger. For this aim, we first designed
and characterized a ring-type wearable device. Then, we con-
ducted psychophysical experiments in which 20 participants
compared the perceived similarity of virtual textures generated
via our device using three different modulation methodologies
(FM, AM, and NM) to their 3D printed real counterparts.
During the experiments, the participants explored the printed
textures via their index fingertips; they felt the virtual ones by
moving their index fingers on a smooth surface or in the air.
Then in another psychophysical experiment, the same partici-
pants compared the similarity of virtual textures rendered via
different modulation types and intensities. They also rated the
perceived realism and pleasantness of the rendered textures.
A. Comparison of virtual textures to their real counterparts
Our findings showed that using frequency modulation (FM)
to render textures with a low spatial frequency significantly
improved their perceived similarity to their real counterparts
(check the first column in Fig. 6). Nonetheless, the modulation
type did not make a perceptual difference when rendering
textures at high spatial frequencies (compare columns in
Fig. 6). Reduced sensitivity to modulation type at higher
spatial frequencies may be due to differences in discrimination
sensitivity at different frequencies [28]. Another reason could
be the distinct mechanisms underlying the perception of coarse
and fine textures. For example, previous research [29] showed
that for coarse textures, both spatial deformation of the finger-
tip and the rate of change of these deformations play a role
in their roughness perception; total vibratory power becomes
more dominant for fine ones [30]. Earlier research conducted
on a surface haptic display [31] showed evidence that due to
these reasons, fine textures could be rendered by considering
only a few highest components in their frequency spectrum,
while coarser ones need more precision. Considering these
studies, for rendering textures with low spatial frequencies,
modifying the frequency of relocated vibrations similar to
the fingertip vibrations might have helped participants better

9
associate them with their real counterparts. However, the
tested methodologies did not cause significant differences in
vibration power, causing indifference to perceptual similarities
for rendering fine textures.
Interestingly, exploring virtual textures on a flat surface hav-
ing the same material or in the air by not making any contact
did not make a significant difference when comparing them
with their 3D printed counterparts (check Fig. 6). This result
was unexpected because when humans interact with surfaces
with their fingertips, they feel not only contact vibrations but
also other properties, such as friction, thermal conductance,
and stiffness, which the participants were deprived of in
contactless conditions. Moreover, there is evidence in the
literature [32], [33] that remote vibrotactile stimulus can alter
the perception of a real texture simultaneously encountered at
the fingertip. The absence of measurable difference between
our two contact conditions, despite these observed differences
in the literature, could be due to several factors. One reason
could be that the participants mainly relied on the vibration
cues and ignored the others during the comparison test [34]. In
fact, earlier studies [30] demonstrated that vibrations generated
during fine texture exploration correlate with roughness per-
ception, and roughness is one of the most dominant perceptual
dimensions [35]. Another potential reason is the inconsistency
across participant preference for one contact condition versus
the other, as revealed by the post experiment questionnaire.
It is worth discussing here the implications of asking our
participants to compare similarity of a real texture, felt on the
glabrous skin of the fingertip, to a vibration applied to the
hairy skin of the distal finger joint. Although not measured
in our study, we expect that vibrations within our actuator’s
frequency range will easily travel the length of the entire finger
to reach the other location; see [36] for a characterization of
frequency-dependent wave propagation on the human hand.
Amplitudes will diminish as vibrations propagate away from
their source, but the single-frequency values used in this study
will remain the same frequency. While these texture sensa-
tions applied at different locations are certainly not identical,
they will engage a large and overlapping area of Pacinian
mechanoreceptors. It would be interesting to observe partici-
pants’ ability to make similarity judgements between texture-
induced vibrations applied to locations with non-overlapping
receptive fields, such as the fingertip and the wrist.
B. Comparison of virtual textures between each other
When the participants compared the perceived realism of
the different virtual textures, they felt the textures rendered via
FM were more realistic than ones via AM and NM for both
spatial frequencies (see Fig. 7). This result demonstrates that
even though the textures were generated at a remote location,
altering the frequency of the vibrations akin to ones occurring
at the fingertip can generate noticeably more realistic textures.
Nonetheless, for FM conditions, using higher amplitudes for
rendering low spatial frequency textures led to more realistic
rendering; however, this situation was the opposite for the ones
with high spatial frequency (compare the pie charts in the
first row in Fig. 8). This phenomenon is not surprising as the
amplitude of the most dominant frequency component of the
finger-contact vibrations that occur during interaction with 3D
printed surfaces is highest at 40 Hz and lowest at 120 Hz (see
Fig. 1).
Participants in aggregate appeared to find any type of
modulation more pleasant than no modulation, across both fre-
quencies and modulation types. Surprisingly, these preferences
were present regardless of similarity ratings; despite FM and
AM textures being rated as less similar than other modulation
pairings, neither stands out as more pleasant than the other, yet
both are consistently more often chosen as pleasant than NM
textures. Participants may have disliked the abrupt changes in
NM textures, and these findings suggest that any modulation
is useful even if it does not enhance realism.
We found that the amplitude of the rendering signal sig-
nificantly affects the perception of virtual textures with both
low and high spatial frequencies. As anticipated, the similarity
ratings for virtual-virtual texture comparisons were lowest
when the amplitude differences were the highest (see Fig. 8).
Amplitude also had an impact on perceived realism across
frequencies; participants preferred higher displacement ampli-
tudes for the lower frequencies and lower amplitudes for the
higher frequencies. This is in line with the force amplitudes
observed in Fig. 3 and the displacements used in Experiment
1. For all but one case, the virtual textures rendered with lower
amplitude signals were perceived as more pleasant compared
to the ones rendered with higher amplitude (compare the pie
charts in the last row in Fig. 8).
C. Limitations and future directions
Virtual texture vibrations used in this study were very
limited in their frequency content. While we chose single-
frequency textures for simplicity and ease of modulation, most
real textures are composed of much richer spectral informa-
tion. It would be interesting to see if observed differences in
virtual texture modulation extend to richer vibrations measured
from real texture interactions. We were constrained from
looking at higher-frequency textures due to a lack of a real
texture that could produce a comparable and distinguishable
high-frequency spectral component. We also could not look
at lower-frequency textures due to the high pass filtering we
used to protect our vibrotactile actuator from DC offsets.
The number of real textures (and their virtual analogs)
in this study was also very limited, primarily due to our
concerns about overall experimental length. Texture compar-
isons quickly become both mentally and physically exhausting,
negatively impacting perceptual acuity and motivating us to
keep our experimental sessions as short as possible. We believe
that three spatial frequencies was the minimum number needed
to observe initial trends in the perception of modulation types
across frequency. However, our small number of textures,
paired with the fact that they are all composed of single
sinusoids, limits extrapolation of this study’s results to all
possible textures. Our current findings motivate future work
exploring a wider range of textural frequency composition.
Our attempts to mitigate the perceptual impacts of amplitude
differences were imperfect; the strength of texture-induced

10
vibrations can vary considerably across participants and even
individual trials, so we chose to simply set all vibration ampli-
tudes to roughly similar ratios as that seen in measurements of
real texture scans from the first author. Unintended differences
in intensity almost certainly played a role in similarity ratings,
particularly between pairs of stimuli that differed in frequency.
Nonetheless, perceptual differences (or lack thereof) between
modulation types within a single frequency and amplitude
demonstrate that amplitude was certainly not the only factor
in perceived similarity.
Finally, the wearable device and tracking system imposed
additional limitations. The actuator was at the upper limit
of usable size, stretching almost the full length of the distal
phalanx for the average participant. This large size does not
impede dexterity or hand closure, as it lies on the back of
the hand while the securing strap wrapped around the finger
was much narrower. However, users with particularly small
hands could find the fit less comfortable if the actuator extends
slightly past the first knuckle. We do not know the perceptual
impact of our actuator applying vibrations laterally to the
skin surface, or if actuation in the direction normal to skin
surface (i.e., pressing into skin) might feel more realistic;
our preliminary exploration with normal-direction actuators
suggests they may result in more natural sensations. Addi-
tionally, when wearing the device, participants were limited to
a small swipe range. This forced participants into somewhat
unnaturally short movements, and future experimental setups
may benefit from a larger workspace where participants may
move more freely.
D. Conclusions
In summary, study participants were significantly more
sensitive to differences in modulation type for the lowest
frequency virtual textures than for those of higher frequency.
For the lowest frequency textures, FM virtual textures were
more similar to their real counterparts and less similar to
virtual textures using other modulation types. In contrast, we
saw no significant difference in similarity to higher frequency
real textures for different modulation types. This suggests that
preserving spatial frequency in texture rendering at finer length
scales may not be necessary, at least for relocated actuation.
This has the potential to simplify significantly signal design
and modulation for relocated vibrotactile feedback in haptic
texture rendering.
ACKNOWLEDGMENT
This work was generously made possible by funding from
the Delft University of Technology. We acknowledge former
M.Sc. student Sophia Huang (Northwestern University) for her
preliminary investigations of amplitude-modulated vibrations
for texture rendering, Michael Wiertlewski for allowing the
laser doppler vibrometer in his lab as well as all the staff
at 3Delft for their excitement and willingness to push their
3D printers to their spatial resolution limit. We also thank the
anonymous reviewers whose valuable suggestions helped us
improve this manuscript.
REFERENCES
[1] Meta, “Meta quest,” https://store.facebook.com/quest, 2022, [Online;
accessed 4-August-2022].
[2] HTC Corporation, “Vive,” https://www.vive.com/us/, 2022, [Online;
accessed 4-August-2022].
[3] HaptX Inc., “Haptx,” https://haptX.com, 2022, [Online; accessed 4-
August-2022].
[4] SenseGlove, “Senseglove,” https://senseglove.com, 2022, [Online; ac-
cessed 4-August-2022].
[5] N. Zamani and H. Culbertson, “Effects of dental glove thickness
on tactile perception through a tool,” in 2019 IEEE world haptics
conference (WHC). IEEE, 2019, pp. 187–192.
[6] D. Gueorguiev, B. Javot, A. Spiers, and K. J. Kuchenbecker, “Larger
skin-surface contact through a fingertip wearable improves roughness
perception,” in International Conference on Human Haptic Sensing and
Touch Enabled Computer Applications. Springer, 2022, pp. 171–179.
[7] S.-Y. Teng, P. Li, R. Nith, J. Fonseca, and P. Lopes, “Touch&fold: a
foldable haptic actuator for rendering touch in mixed reality,” in Pro-
ceedings of the 2021 CHI Conference on Human Factors in Computing
Systems, 2021, pp. 1–14.
[8] Z. Sun, M. Zhu, and C. Lee, “Augmented tactile-perception and haptic-
feedback rings as human-machine interfaces aiming for immersive
interactions,” Nature Communications, vol. 13, no. 5224, 2022.
[9] C. Pacchierotti, G. Salvietti, I. Hussain, L. Meli, and D. Prattichizzo,
“The hring: A wearable haptic device to avoid occlusions in hand
tracking,” in 2016 IEEE Haptics Symposium (HAPTICS). IEEE, 2016,
pp. 134–139.
[10] E. Pezent, P. Agarwal, J. Hartcher-O’Brien, N. Colonnese, and M. K.
O’Malley, “Design, control, and psychophysics of tasbi: A force-
controlled multimodal haptic bracelet,” IEEE Transactions on Robotics,
2022.
[11] M. Sarac, T. M. Huh, H. Choi, M. R. Cutkosky, M. Di Luca, and A. M.
Okamura, “Perceived intensities of normal and shear skin stimuli using a
wearable haptic bracelet,” IEEE Robotics and Automation Letters, vol. 7,
no. 3, pp. 6099–6106, 2022.
[12] H. Culbertson, J. Unwin, and K. J. Kuchenbecker, “Modeling and
rendering realistic textures from unconstrained tool-surface interactions,”
IEEE transactions on haptics, vol. 7, no. 3, pp. 381–393, 2014.
[13] H. Culbertson and K. J. Kuchenbecker, “Ungrounded haptic augmented
reality system for displaying roughness and friction,” IEEE/ASME
Transactions on Mechatronics, vol. 22, no. 4, pp. 1839–1849, 2017.
[14] B. L. Kodak and Y. Vardar, “Feelpen: A haptic stylus displaying
multimodal texture feels on touchscreens,” 10 2022. [Online]. Available:
https://www.techrxiv.org/articles/preprint/FeelPen A Haptic Stylus
Displaying Multimodal Texture Feels on Touchscreens/21341871
[15] A. ˙
Is¸leyen, Y. Vardar, and C. Basdogan, “Tactile roughness perception
of virtual gratings by electrovibration,” IEEE Transactions on Haptics,
vol. 13, no. 3, pp. 562–570, 2019.
[16] R. F. Friesen, R. L. Klatzky, M. A. Peshkin, and J. E. Colgate, “Building
a navigable fine texture design space,” IEEE Transactions on Haptics,
vol. 14, no. 4, pp. 897–906, 2021.
[17] C. Bernard, J. Monnoyer, and M. Wiertlewski, “Harmonious textures:
The perceptual dimensions of synthetic sinusoidal gratings,” in Inter-
national Conference on Human Haptic Sensing and Touch Enabled
Computer Applications. Springer, 2018, pp. 685–695.
[18] C. Gaudeni, L. Meli, L. A. Jones, and D. Prattichizzo, “Presenting sur-
face features using a haptic ring: A psychophysical study on relocating
vibrotactile feedback,” IEEE transactions on haptics, vol. 12, no. 4, pp.
428–437, 2019.
[19] J. Saint-Aubert, “Identifying tactors locations on the proximal phalanx
of the finger for navigation,” in Haptics: Science, Technology, Appli-
cations: 12th International Conference, EuroHaptics 2020, Leiden, The
Netherlands, September 6–9, 2020, Proceedings. Springer, 2020, pp.
42–50.
[20] A. Talhan, H. Kim, and S. Jeon, “Tactile ring: Multi-mode finger-worn
soft actuator for rich haptic feedback,” IEEE Access, vol. 8, pp. 957–966,
2019.
[21] O. Ariza, P. Lubos, F. Steinicke, and G. Bruder, “Ring-shaped haptic
device with vibrotactile feedback patterns to support natural spatial
interaction,” in Proceedings of the 25th International Conference on
Artificial Reality and Telexistence and 20th Eurographics Symposium
on Virtual Environments, 2015, pp. 175–181.
[22] L. R. Manfredi, H. P. Saal, K. J. Brown, M. C. Zielinski, J. F.
Dammann III, V. S. Polashock, and S. J. Bensmaia, “Natural scenes in
tactile texture,” Journal of neurophysiology, vol. 111, no. 9, pp. 1792–
1802, 2014.

11
[23] C. M. Greenspon, K. R. McLellan, J. D. Lieber, and S. J. Bensmaia,
“Effect of scanning speed on texture-elicited vibrations,” Journal of The
Royal Society Interface, vol. 17, no. 167, p. 20190892, 2020.
[24] E. Vezzoli, T. Sednaoui, M. Amberg, F. Giraud, and B. Lemaire-
Semail, “Texture rendering strategies with a high fidelity-capacitive
visual-haptic friction control device,” in Haptics: Perception, Devices,
Control, and Applications: 10th International Conference, EuroHaptics
2016, London, UK, July 4-7, 2016, Proceedings, Part I 10. Springer,
2016, pp. 251–260.
[25] W. McMahan and K. J. Kuchenbecker, “Dynamic modeling and control
of voice-coil actuators for high-fidelity display of haptic vibrations,”
2014 IEEE Haptics Symposium (HAPTICS), pp. 115–122, 2014.
[26] S. J. Lederman and R. L. Klatzky, “Haptic perception: A tutorial,”
Attention, Perception, & Psychophysics, vol. 71, no. 7, pp. 1439–1459,
2009.
[27] A. M. Murray, R. L. Klatzky, and P. K. Khosla, “Psychophysical
characterization and testbed validation of a wearable vibrotactile glove
for telemanipulation,” Presence: Teleoperators & Virtual Environments,
vol. 12, no. 2, pp. 156–182, 2003.
[28] A. Israr, H. Z. Tan, and C. M. Reed, “Frequency and amplitude
discrimination along the kinestheticcutaneous continuum in the presence
of masking stimuli.” The Journal of the Acoustical Society of America,
vol. 120, pp. 2789–2800, 2006.
[29] A. Smith, C. Chapman, M. Deslandes, J. Langlais, and T. MP, “Role of
friction and tangential force variation in the subjective scaling of tactile
roughness.” Experimental Brain Research, vol. 144, no. 2, pp. 211–23,
2002.
[30] S. Bensmaia and M. Hollins, “The vibrations of texture,” Somatosensory
Motor Research, vol. 20, no. 1, pp. 33–43, 2003.
[31] T. Fiedler and Y. Vardar, “A novel texture rendering approach for
electrostatic displays,” in International Workshop on Haptic and Audio
Interaction Design-HAID2019, 2019.
[32] M. Jamalzadeh, B. G¨
uc¸l¨
u, Y. Vardar, and C. Basdogan, “Effect of remote
masking on detection of electrovibration,” in Proceedings of the IEEE
World Haptics Conference (WHC), Tokyo, Japan, Jul. 2019, pp. 229–
234.
[33] R. T. Verrillo and B. G. Calman, “Masking of vibrotactile sensations
from a remote site,” The Journal of the Acoustical Society of America,
vol. 69, p. S23, 1998.
[34] B. Richardson, Y. Vardar, C. Wallraven, and K. Kuchenbecker, “Learning
to feel textures: Predicting perceptual similarities from unconstrained
finger-surface interactions,” IEEE Transactions on Haptics, 2022.
[35] S. Okamoto, H. Nagano, and Y. Yamada, “Psychophysical dimensions
of tactile perception of textures,” IEEE Transactions on Haptics, vol. 6,
no. 1, pp. 81–93, 2012.
[36] B. Dandu, Y. Shao, A. Stanley, and Y. Visell, “Spatiotemporal haptic
effects from a single actuator via spectral control of cutaneous wave
propagation,” in 2019 IEEE World Haptics Conference (WHC). IEEE,
2019, pp. 425–430.
Rebecca Fenton Friesen joined Texas A&M Uni-
versity in 2022 as an Assistant Professor of Me-
chanical Engineering. She received her Ph.D. in Me-
chanical Engineering from Northwestern University
in 2020, followed by a postdoc in the Cognitive
Robotics Department at Delft University of Technol-
ogy. Her work is centered on surface haptic actua-
tion and perception, particularly for wide bandwidth
virtual texture rendering.
Yasemin Vardar is an Assistant Professor at the
Delft University of Technology in the Netherlands.
She earned her Ph.D. in mechanical engineering at
Koc¸ University in Turkey in 2018 and did post-
doctoral research at the Max Planck Institute (MPI)
for Intelligent Systems in Stuttgart, Germany, until
2020. Her research interests focus on human tactile
perception and haptic interfaces. She received the
2021 NWO VENI Grant, 2018 Eurohaptics Best
Ph.D. Thesis Award, IEEE WHC 2017 Best Poster
Presentation Award, and TUBITAK Ph.D. Fellow-
ship; she was selected for the 2019 MPI Sign Up! Career-building Program.
She is currently a co-chair of the Technical Committee on Haptics.