In recent years, immersive technologies consisting of Virtual Reality (VR) and Mixed Reality (MR) are perceiving high interest from various fields. An open challenge in the VR community is to investigate appropriate ways to deliver notifications to users. This paper proposes to use under-clothing wearables to provide vibrotactile feedback on multiple body locations and investigate the generation of haptic textures using wearable devices. Two user studies were conducted to examine the vibrational parameters' effect on perceived experiences. The first study (N = 6) aims to investigate the minimal needed force for the different vibration textures to be perceived by users. In the second study, a human-computer interaction researcher (N = 1) mapped the texture features to physical features, such as softness and sharpness. This study is a first step towards designing multiple haptic textures for wearables that can provide the appropriate experience for a given scenario (e.g., obstacle avoidance and guidance).
In recent years, immersive technologies such as Virtual Reality (VR) and Mixed Reality (MR) have gained interest from various fields, including education, research, communication, medical supply, and entertainment. These technologies provide immersive experiences with a vast degree of freedom (DOF) compared to more conventional technologies.
However, due to the immersive technologies’ high DOF and immersion, two main challenges arise respectively. Due to the immersion, users can lose track of their bodies in the real world (P1), which can trigger accidents, e.g., hitting an obstacle. Second, the high DOF can cause users to experience fear of losing out (P2). Both challenges are open problems in the VR community.
This paper proposes to use under-clothing wearables to provide vibrotactile feedback on multiple body locations. Some of the applications of our proposed devices and haptic feedback texture include guiding users to a designated content (P2) and preventing them from colliding with real-world obstacles (P1).
For P2, previous works have proposed the
addition of the visual effect as an effective
approach. This approach aims to guide users by
adding certain objects, such as arrows and maps,
which generally provide information through the
object’s metaphor that is already established in the
real world. The high accuracy of this approach is
proven by several previous studies [
Although the visual effect approach has high accuracy as a guidance method, the overtness of the objects used in the method, like the examples written above, became a hindrance to the acquisition of the original visual information of the content. New approaches were advocated to further improve the comprehensive performance of the guidance.
This paper proposes to use under-clothing wearables to provide vibrotactile feedback on multiple body locations. In this manner, the system could deliver notifications that do not interfere with the visual content on the headmount displays (HMDs). Hypothetically, this method could be used to tackle P2, that is, this system could be used to guide users to designated content.
Also, previous works adopted vibration
feedback as a new method to provide guidance.
Nonino et al. [
Compared to this method, our approach with wearable devices has the advantage of the number of presentable vibrational feedbacks. The vibration motors mounted on multiple locations of the user’s body could enlarge the granularity of the guidance information.
Regarding P1, there have been efforts that
consist of utilizing MR technology to convey
information about real-world obstacles while
experiencing VR content. Wu et al. [
As a first step towards designing multiple haptic textures for wearables that can provide the appropriate experience for either obstacle avoidance or guidance scenarios, an investigation was conducted to examine the haptic textures using our wearable devices. The results of this experiment are to be a basis for designing the future system of Attention Guidance and Obstacle Avoidance. Two user studies were conducted to study how different vibration parameters impact the haptic experience. In the first study (N = 6), a vibrotactile threshold investigation took place to find out the minimal needed force for the different vibrations to be perceived by users. In the second study, a human-computer interaction researcher (N = 1) was asked to map the texture features to physical features such as softness and sharpness.
In Chapter 2, there will be an introduction to the previous research pieces related to this paper. Chapter 3 will describe the configurations of the vibrotactile under-clothing wearables. Chapters 4 and 5 will each describe the result and discussion of the two experiments conducted to investigate the vibrotactile textures presented by the proposed system of this paper. In the last chapter 6, there will be a discussion about the conceivable applications, feature work, and a conclusion of the overall paper. 2.
Our paper refers to many previous studies conducted. Those studies could be categorized into wearable devices, vibrotactile stimulation, and applications such as attention guidance, and obstacle avoidance.
Studies have recently advocated wearable
devices as a tool to provide haptic feedback. In
previous studies, one of the most common body
locations considered to provide haptic feedback is
the wrist. There are numerous studies conducted
to advocate devices and systems that aim to
present haptic feedback to the user’s wrist. These
devices mostly consist of multiple haptic
feedback presentation parts and a fastener belt that
is meant to be wrapped around the wrist [
The hands and fingers are other locations of
the body wearable device research that has been
ongoing rapidly. Hinchet et al. [
In another previous study, Bhatia et al. [
Also, Al-Sada et al.’s research [
In the research of Yun et al. [
Strohmeier et al. [
Nakagawa et al. [
Jacob et al. [
In research conducted by Liao et al. [
Xiong et al. [
Some studies that adopt the vibrational
approach for attention guidance, point out the
relationship with the studies conducted about
walking guidance for limited vision people [
Many past studies have considered the objects
used in the visual effect AG method. Wallgren et
al. [
Another research conducted by Speicher et al.
[
Compared to these studies, Nonino et al. [
The haptic feedback method advocated by
Nonino et al. [
Also, some studies advocate a
vibrationpresentable belt device for different body parts.
Paolocci et al. [
Though these research pieces pursue providing an efficient AG, they cannot provide AG in a 3DOF situation due to their vibration motor layout.
In the study conducted by Pescara et al. [
In many studies that aimed to provide a guide
for collision avoidance with a real-world obstacle
in VR content, visual approaches were adopted
e.g., Kanamori et al. [
Nakagawa et al. [
A vibrotactile under clothing wearable to
examine the textures of multiple vibration
presentations on multiple body locations with the
structure obtained from Bhatia et al.’s research [
This experiment aims to examine the vibrational stimuli threshold in multiple locations of the body. To carry out our examination, a prototype of the 3DOF AG system was developed, which can provide vibration in three separate locations. As shown in Figure 2, this system is divided into three divisions. The vibration devices, ra
r a e
The vibration device developed is shown in
Figure 3, which consists of a vibration motor, and
two magnets. There are three of these devices in
this system, and in the experiment, each of them
presents vibrational stimuli to a separate location
of the body. This device adopts the structure of
Bhatia et al.’s research [
There are two parameters, rendering method
and duration time, configuring the vibrations.
These parameters were taken from the research
conducted by Yun et al. [
There are four types of rendering methods defined in this paper: Linear, Log, Gaussian, and Square. These rendering methods are a parameter set to designate a single vibration’s input voltage displacement. The graph of the displacement caused by each type of rendering method is shown in Figure 5 and the equations of each rendering method are shown in Table 2. Descriptions of the terms written in Figure 5 are as follows: (1) Output Voltage meaning the output voltage of the microcomputer (2) meaning the designated maximum output voltage for a single vibration (3) T meaning the duration time for the output to reach the maximum value, which equals to the second parameter.
The duration time parameter consists of three values, 0.05ms, 0.1ms, and 0.15ms.
parameter is set to designate a duration time for
the output voltage to reach the maximum value,
which is shown as the T value in Figure 5. The
value, 0.15ms refers to the shortest duration time
Yun et al. [
The control program is constructed with Unity (Version: 2021.3.11f1). This program transmits the key input data obtained on the Unity program to
Uno in a format capable of transmitting
with serial communication. This program
allows the experiment conductor to control the presenting vibration with simple keyboard inputs. The specification of the PC used to build this program is shown in Table 3. (a) Linear (c) Gaussian (b) Log (d) Square
exp {− 0 0 =
| − | (1 −
) | − | < log2 (2 −
) | − | <
The microcomputer, Arduino Uno controls the vibration motor according to the data obtained from the serial communication with the PC with the control program.
Arduino Uno possesses a function that enables it to change its output voltages within the range of 0V to 5V in 256 steps. By using this function, Arduino Uno can control the vibrational stimuli according to the values of the parameters designated. The parameters and their values will be described in 4.3.2 Vibrational Parameters. The specification of the Arduino Uno is shown in Table 4.
In this chapter, using the developed system described in
Chapter 3, an experiment was conducted, which aimed to examine the threshold of vibrational stimuli presented to locations of the body. Regarding the fact that this (a) A diagram of a single trial (ascending) (b) A diagram of a single trial (descending) vibrational parameters, and the body locations to present the vibrations were selected.
In this experiment, the participants will be presented with a series of vibrations with ascending or descending maximum intensity. A series of vibrations is presented in a single trial and the graphs in Figure 6 are the descriptive sketches of an ascending and descending trial. Each of the triangles in Figure 6 represents a Collar Bone
Upper Arm single vibration. In both types of trials, ascending and descending, participants will be instructed to notify the experiment conductor when they notice a shift in the recognition state, meaning that they started or stopped recognizing the presented vibration. Participants will repeat the trials two times each for all the vibration type and body location combinations. There are 12 vibration types (combinations of four rendering methods and three duration times), three body locations, two trial types (ascending and descending), and two trials for each combination, totaling 144 trials for this experiment.
The three body locations are the shoulder,
collarbone, and upper arm. These locations are
obtained from Bhatia et al. ‘s [
The vibration intensity is controlled by the 256-stepped output voltage of a microcomputer. A single step represents about 0.02V of change in the output voltage, due to the maximum output voltage being 5V. As shown in Figure 6, the maximum intensity of the single vibrations in a trial differs from one another. The maximum intensity of the single vibrations varies between 45 and 255 steps of the output voltage, and the contiguous single vibrations’ maximum intensity displacement is equivalent to 5 steps, meaning about 0.1V. The output voltage of the minimum step 45, generated vibration with a minimum intensity perceivable with the finger, which has a higher vibrational perceptibility than the body locations adopted in this experiment. 4.2. Participants
6 participants (aged 21 – 36 years; all male) volunteered to participate in our study. None of the participants have previous experience with under-clothing wearable devices. The participants were asked to wear a long sleeve shirt and a headphone to block the auditory perception caused by the vibration motor. Figure 7 shows the image of a participant during the experiment.
The experiment required three hours to conduct per participant and was divided into two days, one hour and a half per day. For each day, trials with a single trial type, ascending or descending, were presented. Half of the participants were presented with the ascending trial on the first day, and the descending trial on the second day, while the other half of the participants were presented with the opposite order. The order of the 72 vibrations presented in a single day was randomized.
After the 10-minute explanation about the experiment’s background and goal, informed consent was acquired from the participants. Participants were then instructed to put on the vibration devices in the designated locations of their bodies. The specific device locations for each participant were measured by the experiment conductor and informed to the participants before the device mounting phase. During the experiment, participants were also asked to wear a headphone with a high sound-insulating performance to avoid perceiving the actuating noise caused by vibration motors.
Participants will follow these steps in a single trial: (1) the vibrational stimulus presentation starts (2) push the Enter key when they notice a shift in the recognition state.
The average of the obtained threshold value for each combination of the vibration type and body location is listed in Figure 8, and the values are written with the output voltage step of the microcomputer (0 to 255). The cell of the figure is painted with deeper color when the stimulus intensity value is low, meaning when the vibration types the cell represent is perceivable with lower intensity.
Two-way repeated-measures ANOVA was applied to the threshold value to determine which factors had significant effects on the scores. We observed that the rendering method (F (3, 15) =705.4, p< 0.001), duration time (F (2, 10) = 465.9, p < (a) Shoulder (b) Collarbone (c) Upper Arm Figure 8: Threshold of each vibration type and body locations 0.001), and the interaction between them (F (6, 30) =28.6, p< 0.001) had significant effects on the threshold value. Also, the body location had a significant effect too (F (2, 10) = 9.4, p < 0.01).
To further analyze the effects of each parameter, Ryan’s method was applied to the average of the vibrational threshold. Though most threshold values of vibrations configuring the same rendering method (Fs (2, 40) > 8.0, ps < 0.01) or the same duration time (Fs (3, 45) > 17.5, ps < 0.01) showed a significant difference with each other on the same body part, some of them did not. On the collarbone, there was no significant difference between the Linear-0.1 combination and Log-0.1 combination (F (3, 45) = 81.5, p < 0.05), and Linear-0.15 combination and Log-0.15 combination (F (3, 45) = 51.3, p < 0.05). On the upper arm, there was no significant difference between the Linear-0.15 combination and Log0.15 combination (F (3, 45) = 17.5, p = 0.3597417), Gaussian-0.1 combination and Gaussian-0.15 combination (F (2, 40) = 43.291, p = 0.0681987), and Square-0.1 combination and Square-0.15 combination (F (2, 40) = 8.0, p < 0.005). The graphs of the threshold and the significant differences among them are shown in Appendix (a) to (g).
Table 5 shows the average effect size ( 2) of the rendering method for each duration time obtained from each body location. According to the values, the effect size declines at 0.15 by 0.19 from 0.05 and 0.135 from 0.1. Also, Table 6 shows the average effect size of the duration time for each rendering method obtained from each body location. The Square vibrations had the lowest effect size from the duration time. The effect size at Square declines by 0.323 from Linear, 0.274 from Log, and 0.249 from Gaussian.
The results show that most types of vibrations consist of a unique threshold affected by the rendering method and duration time. From the results, the minimum intensity required to present a vibrational stimulus on the shoulder, collarbone, and upper arm was learned. These results would be a basis when designing vibrotactile underclothing wearable applications e.g., attention guidance (AG) and obstacle avoidance (OA). In these applications, the intensity of the vibrational feedback could be designed regarding the priority and referring to the minimum perceivable intensity of each vibration type and its presenting location.
Also, the effect size differences are an element to discuss the features of the rendering methods and the duration times. The results show that the vibrations consisting of the Square rendering method were least affected by the duration time changes, and the vibrations consisting of 0.15 duration time were least affected by the rendering method type.
After the experiment, three out of six
participants commented that they confused the
presented vibration with their heartbeat around
the intensity of their perceiving threshold.
Previous studies investigate vibrational
heartbeat’s effect on human emotion [
In this experiment, participants were seated and wore a headphone with high sound-insulating ability. However, the application of this device hypothetically contains a great amount of visual and audio stimulus which would possibly interfere with the device’s vibrational stimulus, and the effects cannot be neglected. Regarding these factors, experiments with a more practical environment will be valid for further investigation of the vibrotactile under-clothing wearable’s capability and the perceivable texture of the vibrations.
Participants were instructed to wear a thin shirt over an undershirt for this experiment to fasten the device with the magnets. This designation enabled the device to hold its position stably. However, not all users will be wearing light clothes in the conceivable applications. This indefinite condition of the user’s garment could destabilize the device position, leading to a loss of
Log 131.25 vibrotactile feedback presentability. A further improvement of the device structure, such as the size and the type of magnet adopted in the device could be a countermeasure for this issue.
This experiment aims to map the vibration’s texture features to physical features. By conducting this experiment, we aim to explore each vibration’s feature perceived by the users to further investigate the aptitude for future applications.
The experiment was conducted with the cooperation of a human-computer interaction researcher, who is also one of this paper’s authors. We conducted this experiment only with a researcher because it requires more training to classify varied types of haptic textures.
Five vibrations, Linear-0.1, Log-0.1, Gaussian-0.1, and Square-0.1, were presented to the upper arm, and two vibrations, Linear-0.1, Log-0.1, were presented to the shoulder. All the vibrations have a common duration time of 0.1, the second longest duration of the three adopted in the first experiment conducted in Chapter 4. The vibrations adopted in this experiment have an intensity that is five steps stronger than the participant’s threshold of each vibration e.g., presented step 220 intensity for upper arm Gaussian-0.1 vibration which the threshold was 111.25, step 240 intensity for shoulder Log-0.1 vibration which the threshold was 131.25, and step 240 intensity for upper arm Linear-0.1 vibration which the threshold was 135. The thresholds and the presented steps are shown in Tables 7 and 8.
The locations were decided by testing each type of vibration on three locations, shoulder, collarbone, and upper arm. As a result of the comparison, there was no significant difference in the perceived texture between the collarbone and the other two locations, but there was a slight change in the upper arm and shoulder. At the two locations, the perceived textures of Linear-0.1, and Log-0.1 differed slightly, so the mapping was conducted separately.
The presented vibrations were mapped with 11 pairs of adjectives shown in Table 9. The participant was asked to choose a number between 1 through 7, which represents how accurately the presented vibration matches the adjective, while 1 and 7 are the extreme agreement with each adjective and 4 is neutral.
These adjectives were selected from two
studies conducted by Strohmeier et al. [
The participant was presented with the vibrations and was instructed to choose a number between 1 and 7 for each pair of adjectives. The presentation order was random, and the
Linear participant was able to change the answers after perceiving other vibrations.
The result of the mapping is shown in Table 10. For the vibrations presented on the upper arm, Linear, Log, and Gaussian were evaluated to have a very similar texture in five out of 11 categories.
Table 11 shows the mean of the mapping result per adjective pair. Table 12 shows the difference from the mean of the mapping value of each pair of adjectives, and the sum of the difference for each rendering method. When looking at Table 12 it could be said that the most unique texture of the four is the Square with 10.50 points and followed by Log with 7.50 points.
For Linear and Log, the vibrations were presented to the shoulder too. Linear vibration on the shoulder presented a one-point weaker texture compared to the upper arm. The Log vibration on the shoulder presented a one-point lighter, harder, and two-point more aggressive texture than the upper arm. These results could be a basis for designing an application for the vibrotactile under-clothing wearable.
This paper proposes a vibrotactile underclothing wearable to present vibrational feedback to the users of the contents with immersive technologies e.g., VR and MR. The
vibrations presented by our system are configured by four types of rendering methods and three lengths of duration times, resulting to gain the ability to present 12 types of unique vibrations. To investigate the texture feature of these vibrations, we have conducted two experiments, one to explore the stimulus threshold of each vibration on three body locations, and the other to map the perceived texture feature of the vibrations with multiple adjective pairs.
The first experiment showed the stimulus threshold of the 12 unique vibrations on three body locations, shoulder, collarbone, and upper arm. Most of the vibrations resulted to have a significantly different threshold value from each other, caused by both the rendering method and the duration time.
The second experiment attempted to map the texture features of the vibrations by describing each vibration with 11 pairs of opposite adjectives. The results describe the feature of each vibration with numbers from 1 to 7 which shows the agreement score for each adjective. The Square rendering method appeared to be the most unique method, regarding the results of the average differences between the rendering methods.
The results of the two experiments will be a basis for designing the applications of the vibrotactile under-clothing wearable, such as the attention guidance and obstacle avoidance system for VR and MR. However, the stimulus threshold and the map of the textures gained from the experiments of this paper were gained under a certain condition with the participants seated and the visual and audio noises suppressed. Further study would be valid to investigate the effect of the environmental changes e.g., the participant’s posture, visual and audio noises, and the designation of the participant’s clothes.
Also, there was a comment from a participant that point out the difficulty of mounting the underclothing wearable device by themselves. Hypothetically, this issue is caused by the large size and the low fastening capability. The downsizing and the enhancement of the fastening capability could be achieved by substituting the exterior with an original 3D-printed component, and magnets with a superior model. (a): Thresholds analysis by Duration Time (Shoulder) (b): Thresholds analysis by Rendering Method (Shoulder) (c): Thresholds analysis by Duration Time (Collar Bone) (d): Thresholds analysis by Rendering Method (Collar Bone) (e): Thresholds analysis by Rendering Method (Collar Bone) (f): Thresholds analysis by Rendering Method (Upper Arm) (g): Thresholds analysis by Duration Time (Upper Arm)