Content uploaded by Yusuke Yamazaki
Author content
All content in this area was uploaded by Yusuke Yamazaki on Oct 24, 2022
Content may be subject to copyright.
1
Implementation of Tension-based Compact
Necklace-type Haptic Device Achieving Widespread
Transmission of Low-frequency Vibrations
Yusuke Yamazaki, Hironori Mitake, and Shoichi Hasegawa.
Abstract—The haptic sensation of low-frequency vibration
plays a vital role in the music listening experience, but it can
be enjoyed only in certain facilities and environments. Many
haptic devices have been proposed to convey music audio-
induced vibration for various situations. Such devices require
powerful, low-frequency vibration output and transmission over
a wide area. Making such devices small and user-friendly is
difficult, hindering their popularity. To promote haptic devices
for music listening, this paper describes a method for developing
a practical device using motors and a thread and evaluates
this method’s effectiveness. The proposed necklace-type device
is small (about 55×58×15 mm), lightweight (58.5 g), and easy
to wear, making it suitable for use during everyday travel. In
addition, it can transmit low-frequency (20 Hz) vibrations, whose
amplitude exceeds airborne vibration in a nightclub, to a wide
area across the chest and neck, with a total power consumption of
approximately 2 W. Our proposed method will contribute to the
development of practical and high-performance haptic devices
for music listening.
Index Terms—Haptic Display, Wearable Device, Music Hap-
tics, Entertainment, Human–Computer Interaction.
COPYRIGHT NOTI CE
DOI: 10.1109/TOH.2022.3176673 ©2022 IEEE. Personal
use of this material is permitted. Permission from IEEE must
be obtained for all other uses, in any current or future media,
including reprinting/republishing this material for advertising
or promotional purposes, creating new collective works, for
resale or redistribution to servers or lists, or reuse of any
copyrighted component of this work in other works.
I. INTRODUCTION
Haptic sensations play a vital role in enhancing the ex-
perience of music, movies, and other entertainment. Such
sensations occur due to vibrations transmitted through a solid
structure, such as a floor or chair (structure-borne vibration),
or to air pressure changes from the sound source acting
directly on the entire body (airborne vibration). Examples
of the latter include heavy bass from large sub-woofers in
live-music venues, nightclubs, and cinemas; the explosion of
fireworks; and the sounds of guns and bombs during military
maneuvers. However, feeling such sensations is difficult in
Y. Yamazaki and S. Hasegawa are with the Department of Information
and Communications Engineering, School of Engineering, Tokyo Institute of
Technology, Japan, e-mail: {yus988, hase}@haselab.net
H. Mitake is with the Department of Frontier Media Science, School
of Interdisciplinary Mathematical Sciences, Meiji University, Japan, e-mail:
mitake@haselab.net
daily life because of difficulty creating sufficiently loud noises.
To expand the situations in which the haptic component of
music can be enjoyed, various vibration devices have been
designed [1], [2], [3], [4], [5]. In this paper, we focus primarily
on music listening, among other entertainment content.
A. Requirements of haptic stimulation for music listening
Previous studies have shown that low-frequency vibration
plays a vital role in music listening. Hove et al. found
that simultaneously presenting subjects with the acoustic and
low-frequency vibrations of music enhanced their sense of
“groove” with the music [6]. In another study, Merchel et
al. had subjects sit in a seated whole-body vibration device
and evaluated the quality of their experience when listening
to music with and without vibration based on audio signals
[7]. Four genres of music were included in their experiment,
and subjects mostly preferred pop music with strong frequency
components below 20 Hz.
Vibrations should be received over a wide body area to
improve the experience of audio-haptic music listening. Vibro-
tactile sensation exhibits a spatial summation effect, in which
stimulating a larger area of skin lowers the detection threshold
[8]. In our previous research [9], we conducted a questionnaire
on participants’ impression of experiencing such sensations.
The subjects preferred music listening using our prototype
(Fig. 1(c)), which used a thread to transmit vibration to a wide
area, over a linear vibrator that transmitted vibration from a
single point.
Here, we determine the specific amplitude of low-frequency
vibration that should be transmitted to the body for optimal
music enjoyment. In this paper, we assume a representative
situation of music listening near the loudspeakers in a night-
club, a location that allows for pronounced haptic sensation.
In nightclub music contexts, the acoustic pressure typically
reaches 130 dB [10]. We also define “low frequency” as 20
Hz, which is widely considered the lower limit of the audible
range in everyday circumstances. Takahashi et al. [11] showed
that, when exposed to a 20-Hz sine wave with sound pressure
levels (SPLs) of 100, 105, and 110 dB, vibration acceleration
levels (VALs) of about 80, 85, and 90 dB, respectively, were
measured on the subject’s chest. Supposing that the VAL
increases linearly with the SPL, we estimate that the VAL
is 110 dB when the SPL is 130 dB. We converted VAL into
m/s2using the reference acceleration of 10−6m/s2, as stated
in their paper. Therefore, the estimated value was 0.32 m/s2
2
at 20 Hz, and we define this value as a “reference amplitude”
for the minimum amount of vibration required for enjoyment.
B. Problems with using existing vibration devices
Eccentric motors and linear vibrators, often used in general
haptic devices, cannot be used within small haptic devices to
transmit low-frequency vibrations over a wide area. Because
these vibrators generate vibrations via the reaction force when
moving an internal weight (Fig. 1(a)), the vibrator housing
must be large enough to provide sufficient stroke volume (the
internal weight’s movement range) for useful vibrations. In
addition, vibration transfer to the body takes place at the
contact area with the housing, but the transmitted vibration
amplitude is greatly damped by absorption into the skin
[12]. Thus, transmitting substantial vibration requires a large
vibrator or many small vibrators to increase the contact area,
resulting in a larger haptic device.
C. Proposal
In our previous study, we proposed a vibration-generation
mechanism using motors and a thread (Fig. 1(b)) that could
transmit vibration over the user’s torso [9]. We developed a
prototype using this principle, with the thread wound around
the user’s torso (Fig. 1(c)), and evaluated its vibration trans-
mission performance. This study demonstrated that, in prin-
ciple, a small device could transmit low-frequency vibrations
over a wide area. However, the prototype had several practical
problems: the thin thread was troublesome to wind around the
body and uncomfortable to wear, and the actuators consumed
large amounts of electricity.
This paper proposes a design policy and implementation
method for a necklace-type haptic device using our previ-
ously proposed driving principle and evaluates this method’s
effectiveness and the device’s characteristics. The implemented
device is practical for use during everyday travel, such as com-
muting, walking, and riding in a car or on public transport. It
can transmit low-frequency vibration of greater amplitude than
the airborne vibration transmitted by a loudspeaker playing a
130-dB sound.
II. RELATED WOR KS
Much research has been conducted on the transmission of
vibrations to a wide area of the human body to enhance the
experience of entertainment content.
A. Haptic devices incorporating general vibrators
For example, previous studies proposed clothing- and chair-
type haptic devices embedded with several small vibrators
to directly stimulate any part of the body [2], [13], [14],
[15], [16]. Because of their high spatial resolutions, these
haptic devices are suitable for conveying collision sensations
as well as complex information, such as distance or direction,
by stimulating targeted areas of the body. However, small
vibrators cannot produce vibrations with high amplitudes at
low frequencies, as explained in Section I-B.
Fig. 1. Structure of (a) a linear vibrator and (b) our previously proposed
driving principle. The red arrows indicate actuator movement, and the blue
arrows indicate movement driven by the actuator. The blue and green dots
in the top and sectional views in (b) indicate the maximum and minimum
states of the output for a sinusoidal wave input. (c) Our previously proposed
prototype using our driving principle.
Other haptic devices were developed with only a few large
vibrators embedded in the back of a chair or a backpack [1],
[3], [4], [17]. These devices have a low spatial resolution
because they comprise only a few vibrators, but the large
contact area of large vibrators contributes to each vibrator’s
extensive transmission of vibrations to the user’s body. Large
vibrators can also generate higher amplitudes of low-frequency
vibrations than small vibrators because the mass and stroke of
the internal weights are larger. This type of haptic device is
thus suitable for generating low-frequency vibration derived
from heavy bass when watching movies or listening to music,
as mentioned in Section I. However, the size and weight of
large vibrators make such devices less portable.
B. Transmitting vibrations through media
Some studies have proposed different approaches to effi-
ciently transferring vibrations over a wider area. Sakuragi et
al. [18] showed that placing a vibrator on the clavicle is an
effective way to transmit vibrations through bone conduction
In another study, Kurihara et al. [19] used granular Styrofoam
as a medium for transmitting vibrations from a few small
speakers, resulting in a method for distributing vibrations from
a few small vibrators over a wide area. Withana et al. [20]
applied properties of low-frequency acoustic wave propagation
3
in the human body to propose a low-resolution haptic interface.
Their method, using a seated-type device, could direct haptic
sensations to a specific body area, such as the stomach or
head. These approaches involve transmission through an ob-
ject, which requires simplifying the device and increasing the
transmission range of a small number of vibrators. However,
the vibrators used in these studies are still conventional and,
hence, subject to the trade-off between an individual vibrator’s
size and its amplitude capacity for low-frequency output.
C. Generating vibrations using DC motors
Some researchers have proposed using a rotary direct cur-
rent (DC) motor as an actuator. Yem et al. [21] showed that
DC motors are capable of effectively driving low-frequency
vibrations by harnessing the rotational motor’s counter-torque
and using the motor’s rotor as the vibration mass. In similar
research, Gourishetti et al. [22] proposed a relatively inexpen-
sive method, a combination of a DC motor and a rigid stylus,
for providing high-fidelity vibrotactile feedback comparable
to that provided by a relatively expensive voice coil actuator.
Additionally, Minamizawa et al. [23] and Nakamura et al.
[24] used a rotational motor to drive a belt in contact with
a fingertip. The rotational motion of the motor shaft, as
the vibration source, was ultimately converted into a trans-
lational motion of the skin. This process achieved sizable skin
displacement, effectively representing both static forces and
low-frequency vibrations. The skin would detect artificially
generated sensations, such as holding an object or feeling a
collision or traction in a video game. However, these studies
focused on the fingertips, a localized area, rather than directly
addressing our aim to transmit vibrations over a wide area.
III. DESIGN POLICY
This paper proposes a design of a practical haptic device
that can be used in the everyday travel situations mentioned
in Section I-C and, using our previously proposed driving
principle, can transmit low-frequency signals extensively to the
user’s body. The design requirements are as follows (details
provided in the following subsections):
(1) The device can extensively transmit low-frequency vibra-
tion with an amplitude exceeding the “reference ampli-
tude” defined in Section I-A.
(2) The device is portable, simple, and easy to put on; it can
be used immediately in a public space.
(3) The device’s audio noise is low enough such that other
passengers are not bothered in public transport.
(4) The device’s power consumption is low, extending battery
life and reducing heat generation.
A. Transmitting low-frequency vibration extensively with a
compact device
Our previous work [9] demonstrated that our original driv-
ing principle could transmit low-frequency vibrations effec-
tively. The concept of the vibration generation is illustrated
in Fig. 1(b). The bobbin is attached to the motor shaft, and
one end of the thread is tied to the bobbin. The other end
of the thread is wound around the bobbin several times and
then rolled out of the housing. The thread is wound around
the user’s body and then similarly wound and fixed to another
bobbin attached to a motor shaft located on the opposite side.
Note that the material in contact with the body can be another
material as long as the material is attached to the thread wound
around the bobbin. When alternating current (AC) voltage is
applied to the motor, the motor shaft rotates, causing the
thread to move translationally. The angle of rotation (θin
Fig. 1(b)) has no limitation, so the limitation of vibration
stroke (rθ in Fig. 1(b)) is determined by the length of thread
wound by the motor. The moved thread deforms the contact
area, transmitting vibration to the user. The advantage of this
driving principle is that the vibration stroke can be long,
and it is largely independent of the device size because the
volume of the bobbin-wound thread and, hence, the bobbin, is
small. The device can thus generate higher-amplitude, lower-
frequency vibrations than conventional vibrators of similar
size. In addition, the thread contacts a large portion of the
user’s body and transmits vibrations over a wide area without
requiring several vibrators.
B. Shape of the device
To meet design requirement (2), we propose that the device
be a necklace that users can easily wear without adjusting
the length. Thus, the user would not need to adjust their
clothing (e.g., by removing a jacket or loosening a belt, as
when wearing a belt-type device), and the device could be
quickly taken out of a pocket or bag and put on in a public
space. The thread would contact the chest and wind around the
neck, thus transmitting vibration over a wide area. We limited
the device’s weight to that of a lightweight smartphone since
smartphone neck straps have been shown acceptable in the
market. The drive circuit must be small enough to fit in a
standard clothing pocket.
C. Suppressing audio noise
Given the device’s usage contexts, its audio noise output
should be acceptable to other passengers on public trans-
port. The SPL in a running train or bus is 70–80 dBA
[25], [26]. Because the SPL of a normal conversation with
interlocutors one meter apart is about 60 dBA according to
International Organization for Standardization (ISO) Standard
No. 9921:2003, we aim to keep the SPL of device noise well
below that. During development, we found that audio noise
was mainly generated by friction from the moving parts and
magnetostriction from the motor. Thus, we propose using a
low-friction Teflon tube to guide the moving parts and a low-
pass filter in the drive circuit.
D. Reducing power consumption
Reducing the device’s power consumption is critical to
improving its conveniences, such as longer usage time and
lower heat generation. Our previously developed prototype
applied a continuous DC voltage to the motors to generate a
steady-state torque on the thread, entailing continuous power
4
consumption. A steady-state torque is likewise essential in
necklace-type devices to maintain the thread tension that keeps
the thread wound on the bobbin and achieves proper vibration
output. Without this steady-state torque, the thread would
unwind from the bobbin, allowing the motor shaft to wind the
thread around itself through forward or reverse rotation. This
would cause improper vibration output; for example, when
an AC voltage of sin ωt is input, the output vibration would
be |sin ωt|, halving the amplitude and distorting the intended
waveform. To reduce power consumption, we propose provid-
ing steady baseline tension to the bobbin using a rubber cord
instead of applying a DC voltage.
IV. IMPLEMENTATION
This section describes implementations of (1) the vibration
transmitter, (2) the steady tension mechanism, and (3) the
drive circuit that satisfy the requirements listed in Section III.
Fig. 2shows the structure of the developed device, named
the Hapbeat, and its components. The necklace component is
about 55×58×15 mm, with a mass of 58.5 g, which is smaller
and lighter than a lightweight smartphone.
Fig. 2. (a) Components of the Hapbeat. All dimensions are in millimeters,
where ρis the curvature radius of the tube through which the thread (purple
line) and the rubber (red line) run. (b) Tension direction. The green arrow
indicates the direction of the tension applied to the bobbin when the thread is
wound, and the white arrow indicates the tension caused by the rubber cord.
(c) Dimensions of the bobbin (mm).
A. Vibration transmission component
The material of the vibration transmission component
should be rigid enough to consistently transmit the alternating
tension produced by the motor shaft rotation. We selected a
thread made of ultra-high molecular weight polyethylene fiber
(RUNCL PowerBraid, four strands, 95 lb per 0.50 mm). This
thread can transmit vibrations directly to the body, but it is thin
and uncomfortable to wear. Additionally, the part contacting
the skin gets dirty from sweat and other factors, so separating
it from the housing allows for easy cleaning. Thus, for the
contact material, we selected a high-rigidity satin ribbon (6
mm wide, 100% polyester) that connects to the thread with a
plastic coupling.
A thread guide is needed to reduce friction between the
thread and housing and change the direction of the thread
unwound from the bobbin (Fig. 2(a)). Friction between the
guide and thread needs to be low to avoid additional audio
noise, signal distortion, and damage to the thread. Therefore,
we used a Teflon tube (Chukoh Chemical Industries, Ltd.
PTFE tube, inner diameter of 1 mm, outer diameter of 3 mm)
as a guide. To reduce friction, we minimized the curvature of
the guide within the constraints of the small housing, as shown
in Fig. 2(a).
A coreless motor was chosen to provide a low moment of
inertia, allowing for quick repetitions of forward and reverse
rotation to reproduce audio-like AC signals. The coreless
motor used in the Hapbeat has the following characteristics: 3
V rated voltage, 1.1 Ωarmature resistance, 1.92 W maximum
output, 6.8 mN ·mstall torque, 10 ms mechanical time
constant, 0.6 g·cm2rotor moment of inertia, and 110 ×103
rad/s2angular acceleration.
B. Mechanism generating steady-state torque
We used a rubber cord (KawamuraSeichu Co. Ltd, elastic
cord, 1 mm in diameter, white, natural rubber and rayon,
36 mm long), which combines thinness and a low Young’s
modulus, as the spring material to generate steady thread
tension. One end of the rubber cord was fixed to the bobbin
(green circle in Fig. 2(a)), and the other was fixed to the
guide attached to the housing (yellow circle in Fig. 2(a)),
thereby applying counter-torque when the thread was unwound
from the bobbin (white arrows in Fig. 2(b)). We designed
the bobbin to have four flanges to prevent the thread from
being completely unwound when the Hapbeat is worn around
the neck. The guide for the rubber cord is made of the same
material as the thread guide (inner diameter of 1.5 mm, outer
diameter of 2.5 mm).
C. Driver circuit design
The driver circuit should be battery-powered for portability
and capable of driving the motor at the maximum effectual
output with audio signal input. We adapted a single-cell
lithium-ion battery (3.7 V) for the circuit to permit simple
charging with a smartphone charger. A class-D amplifier IC
(Diodes Incorporated, PAM8403) was used to amplify the
input audio signal. However, a general audio amplifier IC alone
cannot drive the motor (rated at 3 V, 1.1 Ωresistance) at the
maximum effectual output (1.92 W) as the motor requires
currents up to 1.93 A. Therefore, we constructed an H-
bridge circuit by connecting the gate terminal of a field-effect
transistor (Toshiba Electronic Devices & Storage Corporation,
TPC8408) to the output of the amplifier IC. Additionally,
to counter the audio noise described in Section III-C, we
implemented a low-pass RC filter (C = 2.2 µF,R=1kΩ,
cutoff frequency 72 Hz), which can easily be switched on and
off with a slide switch. The schematics and frequency response
results are shown in the appendix.
V. EVAL UATIO N
The following four experiments were conducted to evaluate
how well the Hapbeat, as described in Sections III and IV,
5
satisfies the requirements outlined in Sections Iand III. Sec-
tions V-A–V-B explain how we measured vibration transmitted
from the Hapbeat to the user’s body. We show that the Hapbeat
maintains the performance of our previously proposed driving
principle and transmits low-frequency vibration over a wide
area that exceeds the reference amplitude criterion set in
Section I-A. We also evaluate the fidelity of the transmitted
vibration to the input signal from frequency components.
Sections V-C–V-D show how we measured the power-saving
effect of the steady tension mechanism implemented in Section
IV-B and the effect of internal friction generated by this
mechanism. In Section V-E, we assess the audio noise of
the Hapbeat and the effect of a low-pass filter to determine
whether the Hapbeat is appropriate for use on public transport.
In Sections V-F–V-G, we measure the frequency response and
response time of Hapbeat to inform potential users about the
device.
A. Vibration transmission measurement on subjects’ body
1) Subjects: The subjects were six men. Their heights,
weights, body mass indexes (BMIs), and body fat percentages
are listed in Table I. Their weights and body fat percentages
were measured with a body composition meter (Tanita Cor-
poration, innerScan Dual RD-802).
TABLE I
INFORMATION OF SUBJECTS
sub-1 sub-2 sub-3 sub-4 sub-5 sub-6
Height (cm) 185 160 164 165 178 174
Weight (kg) 61.8 47.3 52.4 56.8 61.1 74.1
BMI (kg/m2)18.1 18.5 19.5 20.9 19.3 24.5
body-fat (%) 12.1 12.3 16.9 18.7 18.8 24.3
2) Measurement points: Measurements were taken in three
different areas: chest, neck side, and nape, as shown in Fig.
3. To determine appropriate measurement points applicable to
multiple subjects, we conducted a pilot experiment measuring
transmitted vibration at the location marked by black and
yellow dots in Fig. 3with one subject (sub-6 in Table I).
Those results were used to select the points marked by yellow
dots in Fig. 3for the following reasons:
Overall, we omitted the left side because the measurements
on both sides were similar and tended to decay with distance
from the ribbon and housing. First, in the chest area, we
selected points near the ribbon where the vibration would
be most intense, which we assumed to be important for
perception. Second, we selected points on the clavicle because
Sakuragi et al. showed that the clavicle provides an effective
transmission channel [18]. Third, we chose a column perpen-
dicular to the housing to observe the mixing of vibrations from
the housing and the ribbon. Finally, we selected points 4 cm
away from contact points A and C in Fig. 3to observe the
decay trends under different contact conditions. In the neck
side and nape area, we selected the points near the ribbon for
the same reasons as those for the chest and selected a central
column to observe the damping trend.
3) Vibration signals and measurement: Input signals were
supplied using a function generator (FeelTech, FY6600-30),
Fig. 3. Location of measurement points (black and yellow dots) measured
in the pilot experiment described in Section V-A. The yellow points were
selected using the pilot results. The red arrow indicates the direction of an
accelerometer measured at points on the red line, and the blue arrows indicate
that direction measured at points on the blue line (parallel to the ribbon). The
orange circled letters are reference points in each row, coded as follows: A
is the contact point between the ribbon and the clavicle; B is the midpoint
between A and C; C is the center of the coupling; D is the midpoint of the
housing; E is the point at which the ribbon crosses the shoulder seam line; F
is the approximate center of the cervical vertebrae.
with amplitudes adjusted to consume 0.5, 1, and 2 W with
one motor at a 20-Hz sine wave. Hereafter, the signal types are
expressed as [frequency]-[power consumption]. For example,
20Hz-1W indicates an input signal with a frequency of 20 Hz
and a voltage adjusted for one motor to consume 1 W. In the
Hapbeat, the left and right motors wind and unwind the thread
simultaneously. The vibration transmitted from the Hapbeat
was measured with an analog triaxial accelerometer (NXP
Semiconductors, MMA7361LC, using 6 G mode, module
size of 10×10 mm, module mass of 0.4 g). Signals were
recorded using an oscilloscope (Tektronix Inc, MDO4024C)
at a sampling rate of 10 kHz and a recording time of one
second. To evaluate the amplitude of the transmitted vibration,
we used aRMS in Eq. 1. To compare this amplitude with the
predetermined reference amplitude, we used zRMS in Eq. 1,
which is the root-mean-square acceleration along the z-axis
(perpendicular to the body surface).
aRMS =v
u
u
t
1
N
N
X
k=1
(|xk−µx|2+|yk−µy|2+|zk−µz|2)
(1)
6
TABLE II
FRICTION COEFFICIENT AND YOU NG’S MODULUS OF THE CLOTHING
Clothing Direction Friction
Coefficient
Young’s modulus
(N/mm2)
T-shirt Vertical 0.188 573
Horizontal 0.203 546
Dress shirt Vertical 0.179 17.6
Horizontal 0.178 11.0
zRMS =v
u
u
t
1
N
N
X
k=1
(|zk−µz|2)(2)
µx=1
N
N
X
k=1
xk(3)
where xk,yk, and zkare sampled data for each axis and N
is equal to 10 ×103(points); Eq. (3) is of the same form for
the yand zaxes.
4) Clothing: In line with an everyday usage context, each
subject wore a T-shirt or dress shirt and an undershirt. These
clothes, made by Uniqlo Co., included a men’s U Crew Neck
Short-Sleeve T-Shirt (100% cotton), a men’s Extra Fine Cotton
Broadcloth Long-Sleeve Shirt (100% cotton), and an AIRism
Micro Mesh V-Neck Short-Sleeve T-Shirt (comprising 83%
nylon and 17% spandex). Subjects up to 170 cm tall wore
the medium (M) sizes; taller subjects wore the large (L) sizes.
The friction coefficients and Young’s moduli of the T-shirt
and dress shirt are shown in Table II. Friction coefficients were
measured using a surface tester (Kato Tech Co Ltd, KES-FB4-
A). For each material, a 20×20 cm fabric sample was prepared
and tested six times using a weight of 400 g (for applying
tension on the sample), and the average friction coefficient
was calculated for each. Young’s moduli were measured with
a tensile tester (A&D Co Ltd, RTF-1250) Following the tensile
testing protocol for fabrics in ISO Standard No. 13934-1:2013,
five samples were cut along the straight grain and along the
bias of the T-shirt, and three were cut from the dress shirt. Each
sample was 50×300 mm, and the gripping interval during the
test was 200 mm. Tests were conducted on each sample, and
the average Young’s modulus was calculated for each material.
5) Subjective evaluation: An experiment was conducted to
verify the relationship between participants’ subjective vibra-
tion perception and the measured acceleration magnitudes.
Subjective perception was evaluated by asking participants
to color-code measurement points in a figure (Fig. 3) with
all explanatory numbers and letters omitted and all points
initially colored black. The subjects were asked to color
red where they felt the vibration clearly without needing
to concentrate and blue where they felt the vibration subtly
only when they concentrated on perceiving it. During this
subjective experiment, vibration was continuously present on
the subjects, and the subjects had no time limit to answer.
Although we could not eliminate adaptation to vibrotactile
stimuli, we accept this method as appropriate because the
Hapbeat use case assumes that stimulating music vibration
occurs for a long time; thus, we wanted to evaluate subjective
perception after adaptation occurs. Furthermore, subjects could
hear the audio noise from the Hapbeat, as mentioned in Section
V-E1.
6) Procedure: First, the measurement points and an outline
of the housing were marked on the subject’s skin and clothing
with a pen so that the Hapbeat could be restored to its
initial location if misaligned during the experiment. Next, the
accelerometer was attached to the measurement point with
double-sided tape, with its x-axis parallel to the blue or
red arrows (depending on location) in Fig. 3and its z-axis
perpendicular to the body surface. To maintain the attachment,
an experimenter held the accelerometer cable 20 cm from
the sensor part, taking care not to press the sensor on the
body. At each point, measurements were taken once for input
signals of 20Hz-0.5W, 20Hz-1W, and 20Hz-2W, and then the
accelerometer was attached to another point. This procedure
was repeated for all points in each region (the chest, neck side,
and nape area).
Each subject stabilized their posture during the experiment
by following instructions to sit on a chair with their upper
body perpendicular to the seat surface, lightly press their back
and head against the pole fixed directly behind the chair, look
straight ahead, and exhale and hold their breath during the
measurement. During the transition to the measurement of the
next region, the subjective evaluations were carried out; then,
breaks were taken as needed. This measurement procedure was
conducted for all subjects once for each T-shirt and dress shirt.
7) Results: Results for the 20Hz-1W signal are shown in
Fig. 4as representative of the overall experiment. Note that the
results outside of the pale orange area are for measurements
conducted on clothing; that is, such results do not present
transmitted vibration on the skin. The minimum mean values
of zRMS between subjects were 0.8 m/s2at 20Hz-0.5W, 1.2
m/s2at 20Hz-1W, and 1.7 m/s2at 20Hz-2W (through both
shirts).
8) Discussion of quantitative evaluation: The results show
that zRMS at all measurement points exceeds the reference
amplitude of 0.32 m/s2defined in Section I-A. Thus, the Hap-
beat maintained the performance of our previously proposed
driving principle and could transmit 20 Hz vibrations around
the chest and neck with an intensity greater than airborne
vibration from a loudspeaker playing a 130 dB sound. The
value of aRMS was highest at the neck side, which can be
explained because a larger contact force, which increases fric-
tional force, occurred on the neck side than on the chest, owing
to the neck’s greater curvature. The symmetric driving of the
ribbon counteracted tangential forces on the nape. Contrary
to expectations, the damping on the clavicle did not differ
markedly from other locations. Because the ribbon did not
vibrate the clavicle sufficiently and the relevant measurement
point was on the clothing, the effect of bone conduction did
not play into the system.
Comparing the results revealed that overall acceleration was
greater for the dress shirts than for the T-shirts. This difference
can be explained by the dress shirt having a larger Young’s
modulus and more rigidity than the T-shirt, thereby reducing
the decay in vibration transmission. The measured acceleration
of the housing was also greater for the dress shirt, which
had a lower friction coefficient. We also observed that the
7
Fig. 4. Vibrations transmitted to the subjects (n= 6) and subjective perceptions. The solid red circles represent the magnitude of aRMS values, and the red
translucent area indicates one standard deviation (±SD). Results in the pale orange areas indicate measurements conducted on the skin, and those in other
areas indicate measurements conducted on clothing (the gray area indicates the dress shirt’s collar). The pie charts at or beside each point show the number
of subjects who answered “clearly perceived” in deep blue and “slightly perceived” in light blue. The measurement points near the black stars are addressed
in Section V-B.
soft and easily deformable T-shirt got caught on the housing,
limiting movement. This phenomenon might also occur with
other general vibrators attached to clothes. High variance was
observed on the neck side area of the dress shirt, except near
the ribbon. This could be due to the individual differences
in wrinkling and tensions of the dress shirt fabric due to the
subjects’ differing body shapes (e.g., shoulder width and chest
circumference).
9) Discussion of subjective evaluation: The subjective re-
sults (pie charts in Fig. 4) demonstrate that most subjects could
perceive transmitted vibration near the housing and ribbon at
the neck side and nape. Furthermore, more subjects perceived
the vibration at the distal point (4 or 7 cm from the ribbon or
housing) over the chest area with the dress shirt than with the
T-shirt. This suggests that vibration of the clothing itself may
enhance the user’s perception of vibration intensity.
Although the accelerations were lower near the coupling
and housing compared to the neck side and nape areas, at least
the same number of subjects perceived the vibration near the
coupling and housing. This may have occurred because the
thicker shapes of the coupling and housing compared to the
ribbon may have affected haptic perception. The auditory noise
from the coupling and the housing might also have contributed
to these perceptions (see Section V-E1). Subjects could hear
the audio noise during the experiment, and this auditory
stimulation might have affected their vibration perception [27],
[28]. However, subjects could answer the vibration distribution
even at the distal points that generated no auditory noise. This
result indicates that the subjects could perceive the transmitted
vibration to the extent that they could classify the distribution
and strength of the vibration on the skin.
B. Evaluation of waveform fidelity
We obtained the frequency spectrum of the transmitted
vibration via the fast Fourier transform method. The three axial
components of the acceleration measurements at the neck side
and nape (the measurement point near the black stars in Fig. 4)
were combined into one principal value. The waveform was
ascertained by detecting the periodic peaks to identify and
extract the first two cycles.
1) Results: Of the results obtained, the waveforms and
frequency spectra of cases featuring rich harmonic components
(neck side of sub-2 and nape of sub-1) and a strong 20-Hz
component (both sub-6) are shown in Fig. 5as representative
examples.
8
Fig. 5. Results of Fourier transform. The top graph in each pair shows
acceleration along the principal axis from a 20-Hz sine wave input; the bottom
graphs display results of the single-sided spectral analysis.
2) Discussion: In all cases, the transmitted vibrations com-
prise 20-Hz waves and harmonics. The transmission to the
subject with a relatively high body fat percentage (i.e., sub-6)
consisted mainly of 20-Hz vibrations, while for subjects with
relatively low body fat percentages (i.e., sub-1 and sub-2),
transmissions comprised mainly higher-frequency harmonics.
Although the number of subjects in this experiment was
insufficient to investigate any correlation with body fat, the dif-
ference in frequency composition of the transmitted vibration
on the skin implies that vibration perception probably differs
depending on the subject’s body shape, body composition, and
clothing.
C. Comparing power consumption with and without the
mechanism
To evaluate the power-saving effect of the mechanism
generating steady-state torque, we determined the power con-
sumption of the Hapbeat motor when DC was applied to
generate steady torque. This experiment required the Hapbeat
motor to be removed from the housing, and the motor’s bobbin
only held the thread. The thread was wound three times around
a single bobbin, and the free end was attached to a digital
force gauge (Nidec-Shimpo Corporation, FPG-5) to measure
the tension. A constant-voltage power supply device applied
voltage to the motor and measured the current. The Hapbeat
includes two motors and weighs 58.5 g, so the tension applied
to each motor would be about 0.3 N under regular usage,
which was our measurement target. The applied voltage was
set to 0 V at the start of measurement and gradually increased
less than 0.05 V per second. When the digital force gauge
indicated 0.3 N, the current displayed on the power supply
device was recorded (resolution of 0.01 A). We conducted the
measurement five times.
1) Results: According to every measurement, the motor
drew 0.32 A of DC to generate steady-state torque. Thus,
using the steady-state torque mechanism in the Hapbeat would
reduce the power consumption by 0.11 W per motor (calcu-
lated based on a motor resistance of 1.1 Ωand Ohm’s law
P=RI2= 1.1V×(0.32 A)2= 0.11 W).
D. Comparing the starting current
The static friction due to the implemented mechanism pro-
posed in Section IV-B might prevent the thread from moving
at the start of driving. To investigate the effect of the static
friction, we measured the starting current values with the
mechanism (Hapbeat case) and without (bare motor case). In
both cases, only one motor was driven for measurement. For
the bare motor case, the starting current was defined as the
current when the shaft rotates from a resting state. In the case
of the Hapbeat, the starting current was defined as the current
value when the thread first moves from rest. The Hapbeat was
suspended from a hook, with the accelerometer attached to
the coupling on the driven motor side using double-sided tape.
The accelerometer detected the first movement; its output was
displayed on the oscilloscope. The same power supply device
as described in Section V-C applied the voltage and measured
the current.
1) Results: Over ten measurements, the starting currents of
the bare motor case were all 0.01 A, while those of the Hapbeat
averaged 0.55 A (standard deviation of 0.02 A). Derived from
Ohm’s law, the voltage applied to the motor terminal (1.1 Ω) at
the first movement was 0.61 V in the Hapbeat case. This result
indicates that the drive circuit needs to draw a current greater
than 0.55 A and that the amplitude of the input signal should
exceed 0.61 V to drive the Hapbeat from a static state. Note
that the specific value of the starting current can vary among
individuals; however, this paper does not discuss individual
differences.
E. Audio noise from the Hapbeat
The audio noise from the Hapbeat was measured in a quiet
room (LAeq = 38 dB) with the subject (sub-6) wearing the
T-shirt and the undershirt. A sound level meter (Shenzhen
Wintact Electronics Co., Ltd, GM1356) was used for the
measurements and placed at the same height as the housing,
at 15 cm from the housing. The input signals were sine waves
generated by the oscilloscope, with frequencies based on the
E12 series (1.0, 1.2, 1.5, 1.8, 2.2, 2.7, 3.3, 3.9, 4.7, 5.6, 6.8,
8.2; these values were multiplied by 1, 10, and 100 for a total
of 36 values ranging from 1–820 Hz), and the final signal was
1000 Hz. The signal amplitudes were adjusted so that each
motor would consume 1 W. The audio noises from each signal
were measured for 20 s with A-weighting, and the obtained
mean SPL underwent background noise correction according
9
to ISO Standard No. 1996-1:1982. Measurements were con-
ducted with and without a low-pass filter. Measurements with
the filter were only taken for frequencies of 82 Hz or more
(i.e., greater than the cutoff frequency of 72 Hz mentioned in
Section IV-C). Input signal amplitude was the same in both
cases.
1) Result: The measured SPLs are shown in Fig. 6. The
sound types generated were as follows (judged qualitatively):
•At 1–3.3 Hz, a collision sound between the coupling and
the housing and an abrasion sound between the housing
and the clothing were produced.
•From 3.9–82 Hz, a clattering sound generated by the
housing’s up-and-down movement was recorded, along
with a pattering sound when the ribbon was tensioned.
•From 100–1000 Hz, a magnetostriction sound (similar
to the sound produced when the input signal is played
through a speaker) was generated by the motor.
Fig. 6. Audio noise measurements.
2) Discussion: The measurements show that the audio
noise from the Hapbeat was below 50 dBA with the low-pass
filter. Considering that the SPL in public transport is around
70–80 dBA, the Hapbeat’s audio noise is likely acceptable for
use in such a situation. Applying the low-pass filter reduced
information above 100 Hz, but the output of the Hapbeat was
already weak in that range (as shown in the next experimental
result, Fig. 8). Merchel et al. also supported using the low-
pass filter: they showed that subjective perceptions of the
auditory-tactile music experience are improved when music-
induced vibrations are low-pass filtered [7]. Thus, the effect of
applying the low-pass filter on the experience would be small
or even positive. The audio noise in the 1–3.3 Hz range can
be suppressed by modifying the materials and housing design;
however, this is likely not necessary because music generally
does not draw on this frequency band.
In terms of effect on the Hapbeat user, audio noise from
the housing is unimportant because the music sound and
headphones or earphones mask the noise. This experiment
could not evaluate audio noise from bone conduction. The
perceptibility of such audio noise must be verified, but we
assume that it has no adverse effect on the experience, based
on the study of Sakuragi et al. The authors showed that pre-
senting music vibrations on the clavicle, which is more likely
to produce bone conduction noise than the Hapbeat approach,
had a positive effect on the music listening experience [18].
F. Frequency response of the Hapbeat
We measured the frequency response of the Hapbeat with a
reproducible method using a mannequin made from a public
three-dimensional (3D) model and a gel sheet to represent
human skin. The experimental arrangement is shown in Fig.
7(c)–(f), and the details are as follows. The mannequin 3D
model was “haf020.obj” from the 2003 AIST/HQL Human
Body Size and Shape Database [?]. The model was divided
into sections as shown in Fig. 7(a) and (b), and each section
was 3D-printed (using Zortrax, Z-PLA Pro material, laminate
thickness of 0.29 mm and infill 10% hexagonal). The printed
parts were joined with double-sided tape. The gel sheet (Exseal
Co., Ltd., H0-3K) was sticky on one side and smooth on the
other. The smooth side was covered with a urethane film.
A 50×500 mm piece was attached around the neck of the
mannequin, and a 230×290 mm piece was attached to the
chest part.
Four measurements were taken to obtain accelerations on
the chest, housing, nape, and neck side. To measure up to 1000
Hz, we used another accelerometer with a wide frequency
bandwidth (Analog Devices, Inc., ADXL354C, module size
of 18×12.7 mm, module mass of 1 g). The accelerometer was
applied directly onto the location shown in Fig. 7(c)–(f) using
double-sided tape, and the sensor cable, which was 20 cm
away from the sensor module, was taped to the mannequin.
The input signal was the same as used in Section V-E, and
the output was recorded with the oscilloscope with a sampling
rate of 1 kHz for 1–8.2 Hz inputs and 10 kHz for inputs over
10 Hz.
Fig. 7. Schematic of the dimensions of the 3D-printed mannequin (a)–(b)
and measurement conditions (c)–(f). The letters in (a) and (b) indicate the
dimensions: A= 150 mm, B= 62.5 mm, C= 1,100 mm, and D= 70 mm.
The red and blue lines in (a) correspond to those in (b). The green arrows in
(c)–(f) indicate the direction of the accelerometer’s x-axis.
10
1) Results: Results are shown in Fig. 8. The aRMS was over
10 m/s2when the frequency ranges were 2.2–470 Hz for the
position at the chest, 10–100 Hz at the housing, and 2.2–270
Hz at the neck side, while the aRMS measured at the nape
never exceeded 10 m/s2.
Fig. 8. Frequency response of the Hapbeat on the mannequin.
G. Response time of the Hapbeat
To determine the response time of the Hapbeat, the change
in thread tension was measured with a force sensor (PCB
Piezotronics, Inc., 208C01) when a step voltage was applied
to the motor. The Hapbeat hung on a 3D-printed hook, and
the hook was attached to the sensor tip. Step voltage was
applied by turning on a battery box that contained two 1.2-
V AA cells in series. To normalize the measured data, the
values before the application of step voltages were set to zero,
and the steady-state average after input was set to one. The
steady-state criterion was defined as voltage drift within 0.1 V
per second. The duration of interest was from the normalized
input voltage exceeding 0.9 V to the normalized tension value
exceeding 0.9 N.
1) Results: The average delay over the ten trials was 2.74
ms, with a maximum of 3.00 ms.
VI. GENER AL DISCUSSION AND LIMITATIONS
The experimental results demonstrate that the Hapbeat is a
successful haptic device, employing our previously proposed
driving principle, that can be used in everyday travel situations,
such as commuting or using public transportation. The Hap-
beat is easily wearable (can be put on in less than five seconds)
and battery-powered, which makes it convenient, and its audio
noise is less than 50 dBA with a low-pass filter.
In addition, the Hapbeat can effectively transmit low-
frequency vibration. It can transmit 20 Hz vibrations over the
chest and neck with an intensity greater than airborne vibration
from a loudspeaker playing a 130 dB sound. Furthermore, no
significant attenuation occurs up to 470 Hz, and the response
speed is 2.74 m/s2, which is comparable to that of a coreless
motor or linear vibrator [21]. Thus, the Hapbeat covers a wide
range of low music frequencies and responds to instantaneous
input signals, such as drum sounds, with little delay, making
it a high-performing haptic device for music listening.
However, we cannot say definitively that the Hapbeat can
reproduce actual musical vibrations faithfully. The transmitted
vibrations exceeded 10 m/s2in aRMS near the ribbon, which
is much greater than the reference amplitude. Vibrations in
this range are thus perceived more strongly near the ribbon,
differing from the sensory distribution of a live music expe-
rience. Additionally, while the reference value is useful for
comparison with airborne vibration, attempts to reproduce a
realistic music experience should also consider structure-borne
vibration and the sensation caused by the pressure difference
inside and outside the human body. Thus, we cannot directly
compare the experience of using the Hapbeat with that of
being in a nightclub. Nevertheless, this vibration perception
may still contribute positively to the intensity and enjoyment
of a music experience. Hence, the relationship between the
magnitude and distribution of transmitted vibration and the
subjective evaluation of musical haptic experience should be
investigated.
The low-frequency vibration outputs contained strong har-
monics due to poor reproduction of the input signal (Fig.
5). Because the Hapbeat is suspended, the acceleration of
the housing cannot exceed the housing’s gravitational ac-
celeration while the motor unwinds and loosens the thread.
Thus, the Hapbeat sometimes discards the thread-unwinding
part of input signals while maintaining the thread-winding
part. Furthermore, as indicated in Section V-D, the internal
friction between the thread or rubber cord and the Teflon tubes
impedes thread movement in cases of weak input voltage, and
overcoming that friction threshold leads to a rapid, intense
actuation when the input voltage exceeds a certain value.
These phenomena cause the Hapbeat to reflect the input
signal faithfully only under the following conditions: when the
motor output torque exceeds internal static friction and when
the acceleration of the housing is less than the gravitational
acceleration of the housing (in thread-unwinding movements).
These mechanical characteristics introduce harmonics into the
Hapbeat’s output, as shown in Fig. 5. The extent to which
users can perceive these harmonics of transmitted vibrations
much larger than the threshold in actual usage is unknown,
and this needs to be investigated.
Measuring the transmitted vibration directly on the skin
under the clothing was not possible. We tried to attach the
accelerometer directly to the skin by placing it under the
clothing or cutting a hole in the clothing the size of the
accelerometer and attaching it to the exposed skin. However,
we decided against this approach because the transmission
characteristic from the ribbon to the skin differs from the
actual clothing condition. Therefore, we cannot claim from
the quantitative results shown in Fig. 4that vibrations were
transmitted to the skin under the clothing, but the subjective
results shown in Fig. 4indicate that noticeable vibration
was transmitted successfully for the following reasons: some
subjects could perceive vibrations at the distal points, and
results varied depending on clothing type, indicating that
subjects felt vibration from the vibrating clothing. Although
establishing a method for measuring vibration transmitted
to the skin surface through clothing is beyond this paper’s
scope, further investigation is needed to observe such vibration
11
accurately.
VII. CONCLUSION
This paper presents a design and implementation method
of a necklace-type haptic device that uses our previously
proposed driving principle and is practical in for everyday
travel situations, such as commuting or riding on public
transport. Experiments showed that the Hapbeat, which is
compact, lightweight, and easy to wear, can transmit low-
frequency vibrations over a wide skin area. The proposed
method will be helpful for developing a convenient device
capable of transmitting low-frequency vibrations, a goal that
has been difficult to achieve with conventional haptic devices.
Although this paper focuses on the necklace-type device and
assumes everyday usability, the method can also be applied to
other wearable haptic devices, such as belt- or wristband-type
devices. We hope that our proposal contributes to enhance hap-
tic devices’ performance and usability and creates new ways
to use haptic devices for entertainment, ultimately widening
their range of applications.
ACKNOWLEDGMENT
This work was supported by JSPS KAKENHI Grant Num-
bers JP17H01774, JP20H04220.
REFERENCES
[1] M. Karam, C. Branje, G. Nespoli, N. Thompson, F. A. Russo, and D. I.
Fels, “The emoti-chair: an interactive tactile music exhibit,” in CHI’10
Extended Abstracts on Human Factors in Computing Systems, 2010, pp.
3069–3074.
[2] Y. Konishi, N. Hanamitsu, B. Outram, K. Minamizawa, T. Mizuguchi,
and A. Sato, “Synesthesia suit: the full body immersive experience,” in
ACM SIGGRAPH 2016 VR Village, 2016, pp. 1–1.
[3] “SUBPAC,” https://subpac.com/, accessed: 2022-02-14.
[4] “woojer,” https://www.woojer.com/, accessed: 2022-02-14.
[5] “bhaptics,” https://www.bhaptics.com/, accessed: 2022-02-14.
[6] M. J. Hove, S. A. Martinez, and J. Stupacher, “Feel the bass: Music pre-
sented to tactile and auditory modalities increases aesthetic appreciation
and body movement.” Journal of Experimental Psychology: General,
vol. 149, no. 6, p. 1137, 2020.
[7] S. Merchel and M. E. Altinsoy, “Auditory-tactile experience of music,”
in Musical Haptics. Springer, Cham, 2018, pp. 123–148.
[8] R. T. Verrillo, “Vibration sensation in humans,” Music Perception, vol. 9,
no. 3, pp. 281–302, 1992.
[9] Y. Yamazaki, H. Mitake, and S. Hasegawa, “Tension-based wearable
vibroacoustic device for music appreciation,” in International conference
on human haptic sensing and touch enabled computer applications.
Springer, 2016, pp. 273–283.
[10] “What dB level can you be exposed to in a Nightclub and how dangerous
is it for your hearing?” https://www.instrumentchoice.com.au/what-db-
level-can-you- be-exposed- to-in-a-nightclub-and- how-dangerous-is-it-
for-your-hearing, accessed: 2022-02-14.
[11] Y. Takahashi, K. Kanada, and Y. Yonekawa, “Some characteristics of
human body surface vibration induced by low frequency noise,” Journal
of Low Frequency Noise, Vibration and Active Control, vol. 21, no. 1,
pp. 9–19, 2002.
[12] K. O. Sofia and L. Jones, “Mechanical and psychophysical studies
of surface wave propagation during vibrotactile stimulation,” IEEE
transactions on haptics, vol. 6, no. 3, pp. 320–329, 2013.
[13] P. Lemmens, F. Crompvoets, D. Brokken, J. Van Den Eerenbeemd,
and G.-J. de Vries, “A body-conforming tactile jacket to enrich movie
viewing,” in World Haptics 2009-Third Joint EuroHaptics conference
and Symposium on Haptic Interfaces for Virtual Environment and
Teleoperator Systems. IEEE, 2009, pp. 7–12.
[14] Y. Chandra, B. Tag, R. L. Peiris, and K. Minamizawa, “Preliminary in-
vestigation of across-body vibrotactile pattern for the design of affective
furniture,” in 2020 IEEE Haptics Symposium (HAPTICS). IEEE, 2020,
pp. 671–676.
[15] A. Israr, S.-C. Kim, J. Stec, and I. Poupyrev, “Surround haptics:
tactile feedback for immersive gaming experiences,” in CHI’12 Extended
Abstracts on Human Factors in Computing Systems, 2012, pp. 1087–
1090.
[16] J. H. Hogema, S. C. De Vries, J. B. F. Van Erp, and R. J. Kiefer, “A
tactile seat for direction coding in car driving: Field evaluation,” IEEE
Transactions on Haptics, vol. 2, no. 4, pp. 181–188, Oct 2009.
[17] S. Nanayakkara, E. Taylor, L. Wyse, and S. H. Ong, “An enhanced
musical experience for the deaf: design and evaluation of a music display
and a haptic chair,” in Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems, 2009, pp. 337–346.
[18] R. Sakuragi, S. Ikeno, R. Okazaki, and H. Kajimoto, “Collarbeat:
Whole body vibrotactile presentation via the collarbone to enrich music
listening experience.” in ICAT-EGVE. Citeseer, 2015, pp. 141–146.
[19] Y. Kurihara, M. Koge, R. Okazaki, and H. Kajimoto, “Large-area tactile
display using vibration transmission of jammed particles,” in 2014 IEEE
Haptics Symposium (HAPTICS). IEEE, 2014, pp. 313–318.
[20] A. Withana, S. Koyama, D. Saakes, K. Minamizawa, M. Inami, and
S. Nanayakkara, “Rippletouch: initial exploration of a wave resonant
based full body haptic interface,” in Proceedings of the 6th Augmented
Human International Conference, 2015, pp. 61–68.
[21] V. Yem, R. Okazaki, and H. Kajimoto, “Vibrotactile and pseudo force
presentation using motor rotational acceleration,” in 2016 IEEE Haptics
Symposium (HAPTICS). IEEE, 2016, pp. 47–51.
[22] R. Gourishetti and K. J. Kuchenbecker, “Evaluation of vibrotactile output
from a rotating motor actuator,” IEEE Transactions on Haptics, vol. 15,
no. 1, pp. 39–44, 2021.
[23] K. Minamizawa, S. Fukamachi, H. Kajimoto, N. Kawakami, and
S. Tachi, “Gravity grabber: wearable haptic display to present virtual
mass sensation,” in ACM SIGGRAPH 2007 emerging technologies, 2007,
pp. 8–es.
[24] T. Nakamura, V. Yem, and H. Kajimoto, “Hapbelt: haptic display for
presenting vibrotactile and force sense using belt-winding mechanism,”
in SIGGRAPH Asia 2017 Emerging Technologies, 2017, pp. 1–2.
[25] L. Yan, Z. Chen, Y. Zou, X. He, C. Cai, K. Yu, and X. Zhu, “Field
study of the interior noise and vibration of a metro vehicle running
on a viaduct: A case study in guangzhou,” International journal of
environmental research and public health, vol. 17, no. 8, p. 2807, 2020.
[26] S. Hapuarachchi, D. Jayaratna, C. Kalansuriya, and A. Pannila, “Noise
level survery inside the inter provincial buses in sri lanka,” in Proceed-
ings of the Technical Sessions, vol. 23, 2007, pp. 35–40.
[27] S. Guest, C. Catmur, D. Lloyd, and C. Spence, “Audiotactile interactions
in roughness perception,” Experimental Brain Research, vol. 146, no. 2,
pp. 161–171, 2002.
[28] C. Spence and A. Gallace, “Recent developments in the study of
tactile attention.” Canadian Journal of Experimental Psychology/Revue
canadienne de psychologie exp´
erimentale, vol. 61, no. 3, p. 196, 2007.
Yusuke Yamazaki received the M.S. degree in
information and communications engineering from
Tokyo Institute of Technology in 2017. He es-
tablished Hapbeat LLC. in 2017 to commercialize
haptic device using invented vibration mechanism.
His research interest is the popularization and social
implementation of haptic technology relating to XR
technologies and entertainment.
Hironori Mitake received the D.Eng. degree in
computational intelligence and systems from the
Tokyo Institute of Technology, Tokyo, Japan. He
has been an Associte Professor with the Meiji Uni-
versity since 2022 and was previously an Assistant
Professor with the Tokyo Institute of Technology
since 2011. His domain of research includes virtual
creature, embodied agent, haptic interaction, enter-
tainment computing, and virtual reality.
12
Shoichi Hasegawa received the D.Eng. degree in
computational intelligence and systems from the
Tokyo Institute of Technology, Tokyo, Japan. He has
been an Associate Professor with the Tokyo Institute
of Technology since 2010 and was previously an
Associate Professor with the University of Electro
Communications. His domain of research includes
haptic renderings, realtime simulations, interactive
characters, soft and entertainment robotics, and vir-
tual reality.
