Soft Robotic Tongue Mimicking English Pronunciation Movements

Report: Fabrication and Experimental Evaluation

Evan Krisdityawan

, Sho Yokota

, Akihiro Matsumoto

, Daisuke Chugo

Satoshi Muramatsu

and Hiroshi Hashimoto

Department of Mechanical Engineering, Toyo University, Saitama, Japan

School of Engineering, Kwansei Gakuin University, Sanda, Japan

Department of Applied Computer Engineering, Tokai University, Hiratsuka, Japan

Advanced Institute of Industrial Technology, Shinagawa, Japan

Keywords: Soft Robotic Tongue, Pneumatic Network, Fabrication, English Pronunciation.

Abstract: A novel soft robotic tongue mimicking the movements of English pronunciation was proposed, aiming at the

learning support for English pronunciation. A soft robotic tongue’s system design and actuator arrangements

have been proposed, and the Finite Element Methods (FEM) simulation for each deformation has been

conducted. In this paper, we discussed two milestones: fabrication and experimental evaluation. The

fabrication, molding, and casting method was applied to the model, and it was manufactured five times bigger

than the original size of a human tongue. A silicone rubber Ecoflex 00-30 was utilized and poured into the

mold that was preliminary printed with a 3D printer. Moreover, an experiment was conducted to confirm and

evaluate the deformation patterns of English pronunciation movements. A ruler was used to measure the

parameters in each deformation, such as bend and flap angle, and bulge height. It presented that bend and

bulge deformations between the fabricated soft robotic tongue and simulated FEM results were likely the

same; however, the flap deformation slightly differed in the experimental evaluation.

1 INTRODUCTION

Before we dive into current research, we were

experimenting with the practical use of humanoid

robots to improve Japanese’ English pronunciation

and prosody (Krisdityawan et al., 2022). When the

experiment was done, we received a critical comment

from the participants: they did not know how to move

their tongue in the particular sound of English

pronunciation. Specifically, R (/r/), L (/l/), Th- (θ and

ð), F (/f/), and V (/v/) sound. The humanoid robot we

used does not have a function to move the mouth and

tongue. Therefore, we aim to develop a robotic

tongue to visualize the movements during

pronouncing English words, aiming at learning

support. We considered the universality of the robotic

tongue that we design can be used for non-native

English speakers; however, in this paper, our target is

to support Japanese people learning English

https://orcid.org/0000-0002-8507-5620

https://orcid.org/0000-0002-3004-7235

https://orcid.org/0000-0002-3884-3746

https://orcid.org/0000-0003-2416-8038

pronunciation by showing the visual of tongue

movements.

The tongue movements during pronouncing

English words are aggressive and have a high

flexibility rate. To design a robotic tongue that can

mimic English pronunciation movements, it needs to

determine what kind of concept we want to apply and

assign the tongue materials that will be used.

Compared to other works of tongue robotics, they

mainly develop a structure consisting of rigid

materials or use a skeletal structure (Endo et al., 2020;

Hofe et al., 2008; Marconati et al., 2020; Zheng et al.,

2018; Shijo et al., 2019), and the movements were

stiffed. To increase the elasticity and flexibility, some

papers (Lavoisier et al., 2022; Darmont &

Radhakrishnan, 2021) developed a tongue robot using

the fundamentals of soft robotics. These works used

Ecoflex 00-30 silicone rubber and

polydimethylsiloxane (PDMS) (Lavoisier et al.,

704

Krisdityawan, E., Yokota, S., Matsumoto, A., Chugo, D., Muramatsu, S. and Hashimoto, H.

Soft Robotic Tongue Mimicking English Pronunciation Movements 2 Report: Fabrication and Exper imental Evaluation.

DOI: 10.5220/0012207800003543

In Proceedings of the 20th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2023) - Volume 1, pages 704-710

ISBN: 978-989-758-670-5; ISSN: 2184-2809

2022) as their materials and actuating their robot with

air pressure or “PneuNets” (pneumatic network).

PneuNets is a series of chambers embedded in an

elastic material connected to an inelastic layer, and it

starts to inflate when air is pressurized. Using

PneuNets, our robot can increase the flexibility in

movements and speed close to humans when

pronouncing English words.

Designing our soft robotic tongue should fulfill

the conditions of basic tongue movements in English

pronunciation. We analyzed the basic movements

using software called “Pronunciation Coach 3D”

(icSpeech, n.d.). Basic tongue movements during

pronouncing English sounds are distinguished into

three movements: bend, bulge, and flap shown in

Figure 1. Bend is when the tongue tip starts to lift at

a certain angle and can be found at L (/l/) sound.

Bulge is when the middle part of the tongue begins to

bulge or lift at a certain height, which can be found at

R (/r/) sound. The flap is a movement when both the

left and right parts of the tongue lift symmetrically

and circularly to the middle part, similar to the letter

U or V, found in Th- (θ and ð) sound. All parameters,

such as angle, length, and height summarized in Table

1. A parameter 𝐿



and

𝜃



were revised from 0.12

to 0.21 and 60 degrees to 38 degrees due to typing

(a)

(b)

(c)

Figure 1: Three basic tongue movements during English

pronunciation: (a) bend movement. (b) bulge movement.

error during data acquisition. The parameter was

normalized into 1 to make it easier for us to fabricate

the actual tongue robot with different scales or sizes.

The details of the design description will be discussed

in the next section.

Table 1: Ratio designed parameters of the tongue.

Parameter notation Ratio

𝐿

1.00

𝐿





0.21

𝜃



[deg]

𝜃



[deg]

∆ℎ

0.22

𝐿



0.95

(a)

(b)

Figure 2: Overall soft robotic tongue (a) upper part design.

(b) down part design.

2 PREVIOUS RESEARCH

Ahead of our current progress, we would like to

review our previous research (Krisdityawan et al.,

2023) that will discuss the design of our novel soft

robotic tongue and verify the proposed system using

FEM (Finite Element Methods) simulation.

2.1 Design Description

The soft robotic tongue design that we proposed is

shown in Figure 2. It comprises a base plate and three

soft actuators attached to the base plate that can be

satisfied from three basic movements during English

pronunciation.

Soft Robotic Tongue Mimicking English Pronunciation Movements 2 Report: Fabrication and Experimental Evaluation

705

2.2 FEM Simulation

Based on our proposed design, we conducted

simulations using Ansys Mechanical to verify the

deformation. The simulation flow chart is depicted in

Figure 3. It is divided into seven milestones, and the

details explanation is described in Table 2.

Table 2: Seven milestones and details of the simulation.

No. Milestone Description

1 Model Design 3D model CAD data

made with SolidWorks

and imported to Ansys

2 Assign Material Assign the material which

is hyperelastic material

(Ecoflex 00-30)

3 Insert Constrain Setting up Young’s

modulus, Poisson’s ratio,

and Yeoh 2nd orders

4 Mesh Defining 3D shape with

polygonal representation.

5 Analysis Setting Setting up the analysis

including pressure, earth

gravity, fixed point,

frictionless between

walls, and activate large

deflection

6 Calculation Calculate the simulation

based on the settings and

conditions

7 Result Showing the result of the

FEM simulation

Figure 3: FEM simulation flow chart.

Figure 4 shows the FEM simulation results on each

deformation. All deformations pressure was set up to

100 kPa. Bend deformation shows a bending angle of

10 degrees after pressurizing 15 seconds of constant

pressure. The bulge deformation determined that the

highest point of the bulge actuator could reach the soft

palate, which in our model, would be 38.5 mm from

the normal condition. The flap deformation indicated

39 degrees of flap angle symmetrically.

3 FABRICATION & ASSEMBLE

GUIDELINES

There are seven types of fabrication in the soft

robotics field: molding, reinforcements, additive

(a)

(b)

(c)

Figure 4: FEM simulation result on each deformation. (a)

bend deformation. (b) bulge deformation. (c) flap

deformation.

manufacturing, thin-film manufacturing, shape

deposition manufacturing, bonding, and architectural

considerations (Schmitt et al., 2018). In our case, to

produce a soft actuator (PneuNets actuator) which is

the internal structure is crucial, we are applying the

molding (molding and casting) method. Molding and

casting methods are easy to utilize for the soft robotic

field, and the cost is reasonable.

Our soft robotic tongue required preliminary

preparation and multiple processes before it

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

706

assembles into one system. We have tried many

fabrications, and copying the actual size of the human

tongue is challenging to produce due to the small

scale of the air inlet, and some chamber layers were

ripped even before demold or separated from the

mold. Accordingly, we determined to fabricate all

parts into five times bigger than the original size of a

human tongue. Based on the average of Oliver’s

measuring results (Oliver & Evans, 1986), the human

tongue has a dimension: length of 34.95 mm, breadth

of 43.70 mm, and thickness of 10.60 mm. The five

times fabricated size will have a dimension: length of

174.75 mm, width of 218.50 mm, and thickness of

53.00 mm.

3.1 Mold

Mold is needed to copy the model of the soft robotic

tongue. Mold data was made using SolidWorks and

preliminary printed using a 3D printer. The base plate

and each soft actuator mold are shown in Figure 5.

Each mold was sprayed with a release agent before

starting casting.

Figure 5: Base plate (A), bend actuator (B), bulge actuator

(C), and flap actuator (D) mold printed using a 3D printer.

3.2 Casting

For casting, we use Ecoflex 00-30 silicone rubber

liquid with a durometer of 30A produced by Smooth-

On. The silicone liquid consists of Part A and Part B,

which must be mixed before use. Afterwards, the

liquid mixture is injected into the printed mold and

cured for 2-3 hours. In addition, a different color

pigment was added to the mixture to distinguish each

soft actuator in the final product. Figure 6 shows the

casting process.

Figure 6: Scheme of the fabrication process.

3.3 Result

The fabrication result is shown in Figure 7. The

yellow, red, and blue represent bulge, bend, and flap

actuators.

(a)

(b)

Figure 7: Final design of the fabricated soft robotic tongue.

(a) upper part. (b) down part.

4 EXPERIMENTAL RESULTS

This section will discuss the experimental equipment

and method to measure and calculate parameters. In

the experiment, we focus on the measurement to find

the angles (bend and flap) and bulge parameters.

4.1 System Configuration

The equipment we used for the experiment is listed as

follows: one PC, one Arduino, one air compressor,

and three pressure control valves. Figure 8 depicts the

system configuration for the experiment. Electro-

Soft Robotic Tongue Mimicking English Pronunciation Movements 2 Report: Fabrication and Experimental Evaluation

707

pneumatic regulators are used to control the pressure

and act as a continuous process. Each electro-

pneumatic regulator receives the target pressure from

the Arduino and controls the pressure being

pressurized to the actuators. When the target pressure

is issued, we achieve the steady state of the all-

deformation’s soft robotic tongue, which will be

discussed in session 4.2.

Figure 8: System configuration for experimental

verification.

4.2 Method and Results

We configured the air pressure with a 12 kPa flow

every second. Figure 9 shows the experimental results

of the soft robotic tongue in a steady state. For bend

movement, it successfully bent when the air was

pressurized. Bulge movement shows it bulged

vertically and shrank horizontally. While the flap

movement was flapped closed to a letter U. The time

scale for bend, bulge, and flap deformation to reach a

steady state are 0.7 s, 0.8 s, and 15 s, respectively. The

numerical value of experimental and FEM simulation

data represents in Table 3.

(a)

(b)

(c)

Figure 9: Experimental result of each deformation with

comparison before and after pressurized. (a) bend

deformation. (b) bulge deformation. (c) flap deformation.

Table 3: Comparison of experimental and FEM simulation

data.

Deformation

Parameter FEM Data

Experimental

Data

Bend

𝜃



[deg]

𝐿



[mm]

Bulge

∆ℎ [mm]

38.5

𝐿







[mm]

166

164

Flap

𝜃



[deg]

Figure 10 shows the notation parameter and its

presentation using a simplified triangle diagram. The

calculation of

𝜃



and 𝜃



was

derived using tan

-1

While ∆ℎ can be simply measured using a ruler.

(a)

(b)

(c)

Figure 10: Notation parameter during air was pressurized.

(a) bend deformation parameters, (b) bulge deformation

parameters notation. (c) flap deformation with a

presentation of a triangle diagram.

5 DISCUSSIONS

To evaluate the data we obtained from the

experiment, we have to compare the data with the

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

708

simulated FEM data. Based on the summarized data

in Table 3, some experimental data values differed

from the simulated FEM data.

The bend deformation shows a difference of 8

degrees and 8 mm on

𝐿



. For bulge deformation, it

has a slight difference, 0.5 mm in

∆ℎ and 2 mm in

𝐿



. Regarding flap deformation, the flap angle

shows a difference error of 4 degrees. In addition, the

result of flap deformation was not equal

symmetrically when air was pressurized. The left part

of the robot was lifted circularly stronger than the

right part, resulting in the entire system tilted to the

right. It is assumed that there was a human error

during fabricating of the flap actuator; for instance, an

error designed of the flap actuator’s inner chamber

wall (width of the wall) affected the flap actuator

inflating strongly in the radial direction. To solve this,

we plan to redesign and recheck if the design is

identical to the model we used in the FEM simulation.

However, our goal is to use the soft robotic tongue

to support visualizing the tongue motion of English

pronunciation. Even if it has differences in numerical

data, the practical usage is valid overall as long as the

soft robotic tongue can mimic the movements

visually.

6 CONCLUSION & FUTURE

WORKS

We have suggested a novel soft robotic tongue that

mimics English pronunciation movements. We make

it to aim the improvement at learning support for

English pronunciation. This paper discusses the

fabrication process and evaluates the experiment,

including the parameter of tongue movements of

basic English pronunciations. Previously, we

conducted a FEM simulation to confirm the

deformation patterns of each basic tongue movement.

We tried to fabricate the robot and demonstrate the

motion based on the simulation results.

We utilized an elastic and flexible material,

Ecoflex 00-30 elastomer. The fabrication process

starts with making a mold of each soft actuator. After

the fabrication finished, an experiment to validate the

motion was directed. It shows that the movements in

the simulation or the actual fabricated model have

slight differences. The simulated model and

experimental verification contribute to a study of the

soft robotic and soft robotic tongue that mimics

English pronunciation movements.

In the future, it would be better to measure each

parameter using further precise tools such as flex

sensors as feedback to obtain the experimental data

more accurately. Some papers applied flex sensors to

measure the bend or curvature angle in their work

(Roy et al., 2015; Coral et al., 2015), and it gives us

an idea to apply flex sensors to our soft robotic tongue

in the future. Dynamics states analysis will be

conducted in the future for parameter control of each

deformation. Moreover, we will try to fabricate a

tongue cover to cover up the entire system so it does

not bother the appearance for practical usage.

REFERENCES

Coral, W., Rossi, C., & Martin, I. P. (2015). Bio-inspired

morphing caudal fin using shape memory alloy

composites for a fish-like robot: Design, fabrication and

analysis. In 2015 12th International Conference on

Informatics in Control, Automation and Robotics

(ICINCO), Vol. 2, pp. 336-343.

Darmont Araya, F., & Radhakrishnan, P. (2021).

Investigating the Design and Manufacture of PneuNet

Actuators as a Prosthetic Tongue for Mimicking

Human Deglutition. Journal of Engineering and

Science in Medical Diagnostics and Therapy, 4(4).

Endo, N., Kizaki, Y., & Kamamichi, N. (2020). Flexible

pneumatic bending actuator for a robotic tongue.

Journal of Robotics and Mechatronics, 32(5), 894-902.

Hofe, R., & Moore, R. K. (2008). Towards an investigation

of speech energetics using ‘AnTon’: an animatronic

model of a human tongue and vocal tract. Connection

Science, 20(4), 319-336.

icSpeech. (n.d.). Pronunciation Coach 3D.

https://icspeech.com/pronunciation-coach-3d.html.

Krisdityawan, E., Yokota, S., Matsumoto, A., Chugo, D.,

Muramatsu, S., & Hashimoto, H. (2022). Effect of

Embodiment and Improving Japanese Students’

English Pronunciation and Prosody with Humanoid

Robot. 15th International Conference on Human

System Interaction (HSI), (pp. 1-6).

Krisdityawan, E., Yokota, S., Matsumoto, A., Chugo, D.,

Muramatsu, S., & Hashimoto, H. (2023). Soft Robotic

Tongue that Mimicking English Pronunciation

Movements. In 2023 IEEE International Conference on

Mechatronics (ICM), pp. 1-5.

Lavoisier, A., Avila-Sierra, A., Timpe, C., Kuehl, P.,

Wagner, L., Tournier, C., & Ramaioli, M. (2022). A

novel soft robotic pediatric in vitro swallowing device

to gain insights into the swallowability of mini-tablets.

International Journal of Pharmaceutics, 629, 122369.

Lu, X., Xu, W., & Li, X. (2017). A soft robotic tongue—

mechatronic design and surface reconstruction.

IEEE/ASME Transactions on Mechatronics, 22(5),

2102-2110.

Marconati, M., Pani, S., Engmann, J., Burbidge, A., &

Ramaioli, M. (2020). A soft robotic tongue to develop

solutions to manage swallowing disorders. arXiv,

preprint arXiv:2003.01194.

Soft Robotic Tongue Mimicking English Pronunciation Movements 2 Report: Fabrication and Experimental Evaluation

709

Oliver, RG., Evans, SP. (1986). Tongue size, oral cavity

size and speech. Angle Orthod. 56(3):234-43

Roy, K., Idiwal, D. P., Agrawal, A., & Hazra, B. (2015).

Flex sensor based wearable gloves for robotic gripper

control. In Proceedings of the 2015 Conference on

Advances in Robotics, pp. 1-5.

Schmitt, F., Piccin, O., Barbé, L., & Bayle, B. (2018). Soft

robots manufacturing: A review. Frontiers in Robotics

and AI, 5, 84.

Shijo, R., Nagano, M., Mizoguchi, I., Ando, T., Tamura, K.,

Hiki, K., ... & Nojima, T. (2019). Licker: A Tongue

Robot for Representing Realistic Tongue Motions.

SIGGRAPH Asia 2019 Emerging Technologies, pp. 27-

28.

Zheng, Z., Wang, K., Wu, X., Zhang, Q., & Li, K. (2018).

Dynamic analysis of elastic projecting robot inspired by

chameleon tongue. IEEE International Conference on

Robotics and Biomimetics (ROBIO), pp. 2088-2094.

ICINCO 2023 - 20th International Conference on Informatics in Control, Automation and Robotics

710