A Study on the Design Procedure of Three-Dimensional Printable Tactile Graphics for Individuals With Visual Impairments

Abstract

Introduction: Three-dimensional (3D) printing can be used to produce tactile teaching materials for individuals with visual impairments. Although 3D and 2.5D teaching materials are based on realistic content, no systematic conversion model for 2D materials exists. Here, we combined the previous findings and this study's results and technology to develop procedure recommendations for tactile-graphic design. Methods: We enrolled 19 participants with congenital blindness to identify and name the tactile graphics. The tactile graphic design involved height differences, materials, and operation zone areas. We recorded the identification time, accuracy, and National Aeronautics and Space Administration–Task Load Index assessment performance and conducted a three-way analysis of the variance to investigate the interactions. Results: A larger area resulted in a better performance in seconds; model 3 (high lines and low planes) under large areas obtained the lowest speed and accuracy performances. Large areas were associated with a better intelligent load, physical load, mental load, and self-performance scores. Model 1 (equal line and plane heights) yielded better physical-load performance. Model 3 was found to be superior when thermoplastic elastomers were used under large areas. Discussion: Larger area ratios could increase the identification performance. As model 3 is not recommended, model 1 was found to be better than model 4 (non-operation-guiding zones become planes) in terms of physical load. Model 2 (low lines and high planes) required more time than model 1; hence model 1 is recommended. Fused deposition modeling (FDM) materials such as FDM_TPE (thermoplastic elastomer) increases the identification steps and should not be used first. The FDM_PLA (polylactic acid) platform is inexpensive and easier to operate than stereolithography (SLA) materials such as SLA_ABS (acrylonitrile butadiene styrene); hence FDM_PLA is recommended. Implications for Practitioners: When designing the graphics, a plan view photograph of the object can be obtained for delineating the contours first. Adjustments should be made after confirming the presentation model. The graphic may then be transferred for 3D printing with PLA materials.

Keywords

congenital blindness 3D printable tactile graphics tactile-graphic design procedure tactile graphics

Three-dimensional (3D) printing is a promising new technology that enables the implementation of innovative teaching methods and is gradually being integrated into mainstream education (Ferracane, 2020). Since 3D printing is customizable, inexpensive, and can enhance the understanding and memory during the learning process (Butler et al., 2021; Gual et al., 2015; Holloway et al., 2018; Wonjin et al., 2016), studies have started examining the application of this technology in education, and related studies for individuals with visual impairments and have been successively conducted and published (Buehler et al., 2016).

According to our literature review, the existing teaching materials for individuals with visual impairments can be classified into three categories. Brulé et al. (2018) first constructed numerous realistic 3D objects (such as houses and trees) to strengthen the recognition of environmental connections in individuals with visual impairments. Secondly, Cavazos Quero et al. (2021) converted artworks into reliefs (2.5D) and used the length, width, depth, texture, and interactive voice elements to strengthen exploration independence. Further, Stangl et al. (2015) and Panjatevakupt (2020) allowed individuals with visual impairments and their teachers to modify the teaching materials and provide explanations to improve the learning experience.

Since the contents of most 3D and 2.5D teaching materials are realistic, there is a basis for production, and a 3D scanner can be used to obtain 3D images for revision and printing (Elkhuizen et al., 2019). However, the object contours must be simplified when producing 2D teaching materials (only the height of one axis is increased). Holbrook and Andonova (2006) and Carfagni et al. (2012) proposed simplification examples, mainly involving the extraction of the outer contours of the object before different components are combined with different textures. However, they did not include the effects of other tactile-graphic constituent elements.

Therefore, a previous study (Wu et al., 2022) added scale and complexity to the viewing angle determination and representation factor proposed by Holbrook and Andonova (2006) for integrated examination of tactile-graphic constituent elements. Among these elements, it was found that the presentation of medium-scale graphics (palm-sized) as well as mixed lines and planes resulted in a good performance. Accordingly, we employed 3D printing in this study to further examine the effects of various height differences and printing materials on the identification performance.

Regarding height difference, Shimizu et al. (2000) found that the identification performance was better for outline and plane surfaces than for hollowed plane surfaces, while McCallum et al. (2006) found that increasing the height of plane surfaces in one direction does not increase the identification performance as there is no 3D change. 3D printing, microencapsulation, thermoforming, and laser engraving can be used to produce tactile maps for material elements, but the identification performance of 3D printing is greater (Gual et al., 2015). Existing 3D printing techniques include fused deposition modeling (FDM) and stereolithography (SLA). FDM mainly uses 3D printer nozzles to add molten material layer by layer on the platform for stacking to form 3D shapes. The SLA technique uses three-dimensional light for the solidification of photopolymers to form 3D shapes (Cekic et al., 2019). However, Urbas et al. (2016) found that the use of the FDM technique for printing results requires post-production on rough surfaces. Presently, few studies have compared different 3D printing materials, and this was examined in this study.

Conversely, our team has previously examined constituent elements affecting the recognition of tactile graphics to understand the interactions among scale, representation, and complexity factors. Regarding large-scale graphics (larger than a palm), subjects previously mentioned that since the area ratios of some operation prompt areas in the picture were too small, they could not enhance identification (Wu et al., 2021, in press). Therefore, an area variable was added at the same stage in this study.

In summary, an identification experiment was performed for three variables (height differences, material, and operation zone areas). Subsequently, variable analysis was used to examine the objective operation performance of variables in time in seconds and accuracy rate, and the subjective National Aeronautics and Space Administration–Task Load Index (NASA-TLX) assessment performance was analyzed. NASA-Task scale was proposed by Hart and Staveland (1988) and mainly uses multiscale grading for participants to carry out self-assessment of six factors, namely intelligent load, physical load, mental load, self-performance, time load, and frustration. Except for the items pertaining to self-performance, the smaller the value, the better the result. Subsequently, we combined the results of this stage with those of the previous stage to propose a recommended procedure for 2D tactile-graphic design to provide a reference for teachers who need to convert 2D graphic cards using 3D printing.

Methods

This study was conducted in cooperation with the Taipei School for the Visually Impaired and the Office of Disability Services-Resource Room Tamkang University in Taiwan. The study was approved by the Research Ethics Committee of the National Taiwan University Institutional Review Board, and the adult participants and the parents of minor participants signed informed consent before the initiation of the study.

The three variables in this study (height differences at four levels, materials at three levels, and operation zone areas at two levels) resulted in 24 tactile-graphic cards. The participants were required to use both hands for tactile identification and naming of every card. The time needed for the identification, whether the identification was correct, and the NASA-TLX workload assessment score were examined with a three-way analysis of the variance (ANOVA performed with SPSS Statistics v. 20; IBM Corp., Armonk, NY) to identify the best performance combination and summarize the procedure recommendations for designing tactile graphics.

Variable Definition

Height Difference

3D printing can produce various height differences. In this study, we proposed four presentation modes (see Figure 1). Model 1 had equal line and plane heights (plane and line heights, 1 mm), model 2 involved low lines and high planes (plane height, 2 mm; line height, 1 mm), and model 3 involved high lines and low planes (plane height, −1 mm; line height, 1 mm), which are all methods for presenting planes in operation-guiding zones; model 4 involved equal line and plane heights (plane and line heights: 1 mm) but utilized a method for presenting the planes in non-operation-guiding zones.

Figure 1.
Four Models.

Materials

Acrylonitrile butadiene styrene (ABS), polylactic acid (PLA), and thermoplastic elastomer (TPE) are common 3D printing materials (Mardis, 2018). Printing can be conducted using SLA and FDM. SLA produces materials with a hard but smooth feel, while FDM produces materials with a rough texture due to print tracks. Three techniques were used for this variable; printing ABS materials using SLA technology (SLA_ABS); second, printing all three materials using FDM technology (FDM_PLA), and using FDM_TPE. TPE materials are soft and elastic.

Operation Zone Area

The operation zone area mainly refers to the part of the tool an individual would operate with their hands. Two graphic items were used in this experiment, namely a pair of scissors and a compass. The area ratio of the operation zone and the non-operation zone of the pair of scissors is comparable, and it was accordingly classified as large. The sum of the operation zone areas in the compass is much smaller than the non-operation area, and it was accordingly classified as small (see Figure 2).

Figure 2.
Operation Zone.

Participants

We asked the teachers at a school for students with visual impairments to recruit 19 participants without other disabilities (10 male, 9 female; mean age: 23 years). All participants had been taught braille and tactile graphics starting in elementary school. All participants had congenital blindness and had been completely blind before entering the school at the age of 5 years.

Equipment

The base plate was A5 (148 × 210 mm) and had a thickness of 2 mm. The results of the previous study (Wu et al., 2021, in press) were used as a reference for the graphic presentation mode, and the line-and-plane mixed modes with relatively good performance combinations were used for presentation. The operation zone was presented as a plane, while other regions were presented as lines. Line width was set at 3 pt (1 mm) according to Jehoel et al. (2009), and line height was set at 1 mm. SLA_ABS, FDM_PLA, and FDM_TPE were combined with objects with four types of height difference presentation modes and two different operation zone areas to produce 24 graphics cards. The graphics were delineated using a computer based on the actual object before it was imported into the 3D software for closed region extrusion (upward or downward) and stored as an STL file for 3D printing.

Procedure

Four training graphic cards were used for the pretest, and 24 tactile-graphic cards were used for the official experiment, for a total of 28 tactile-graphic cards. In the experiment, a long table (on which two cameras were set up at different angles to record the entire experiment), two chairs, and stationery for recording the experimental results were used.

The experiment was divided into four steps.

Step 1

Four tactile-graphic cards were provided (the scales of models 1 to 4 were presented). An explanation of how inner and outer contours are converted to mixed lines and planes was provided before the definition of the operation-guiding zone was explained, and the differences among the four models were described.

Step 2

The contours, operation instructions, and operation sites of the two objects were explained. The next step started only after confirming that the participants could correspond the object names to the aforementioned explanations. For example:
Scissors consists of two ovals (handles) and two long rectangles (blades). During operation, the thumb and the index finger should be placed in the holes in the oval handles. The fingers should repeatedly bend inward before opening the blades to cut.

Step 3

Graphic cards were randomly arranged. The participants had to name the object after touching every card (subjects must touch and name a card before they are allowed to touch the next card). There was no time limit. If the participants were unable to recognize the graphic, they could state “unable to identify.” The identification results and time needed to identify the graphic were recorded during this process.

Step 4

After the naming was completed, the NASA-TLX assessment was administered via an interview. Using the intelligent load test as an example, we asked the subject whether they needed to focus to remember and identify this graphic. Does not require extreme focus was assigned a score of 1, and extreme focus was assigned a score of 7. A total score scale ranging from 1 to 7 was used for the subjects to select the score.

Results

We conducted a three-way ANOVA to examine whether there was a difference in the scale representation and overall performance of complexity. If there were interactions between factors, a one-way ANOVA was further conducted for a more detailed analysis. The following are the results of the ANOVA for identification time (from card touching to object naming) and accuracy.

Identification Time

The results of the three-way ANOVA for identification time are shown in Table 1. The operation zone area factor was statistically significant among the three factors. The mean time for large areas was 7.6 s, and for small areas was 11.1 s. The difference in the identification time for large and small areas was significant (p = .002). Interactions between operation area ratio and height difference were present (p = .017). When the area was large, significantly less time was needed for model 1 (6.0 s) than for model 3 (10.8 s; p = .016). Moreover, significantly less time was needed for model 4 (6.8 s) than for model 3 (10.8 s; p = .018).

Table 1.
Variable Analysis Results for Performance (Time in Seconds) for Material, Operation Zone Area, and Height Difference.

Variable SS df MS F p

Material 365.320 2 182.660 1.494 .238

Operation zone area 1414.055 1 1414.055 13.607 .002

Height difference 202.059 3 67.353 0.510 .677

MaterialArea 663.741 2 331.871 1.649 .207

MaterialHeight difference 549.224 6 91.537 0.822 .555

AreaHeight difference 960.989 3 320.330 3.688 .017

MaterialAreaHeight difference 451.083 6 75.181 0.723 .632

Note:* df = degrees of freedom; MS = mean square; SS = sum squares.

Accuracy

The results for accuracy are shown in Table 2. Interactions between the operation area ratio and height difference were present (p = .007). When the area was large, the accuracy rate of model 1 (93%) was significantly higher than that of model 3 (75%; p = .047), the accuracy rate of model 2 (97%) was significantly higher than that of model 3 (75%; p = .024), and the accuracy rate of model 4 (95%) was significantly higher than that of model 3 (75%; p = .030).

Table 2.
Variable Analysis Results for the Accuracy Rate for Material, Operation Zone Area, and Height Difference.

Variable SS df MS F p

Material 0.083 2 0.042 0.562 .575

Operation zone area 0.219 1 0.219 1.670 .213

Height difference 0.474 3 0.158 1.616 .196

MaterialArea 0.057 2 0.029 0.552 .581

MaterialHeight difference 0.303 6 0.050 0.685 .662

AreaHeight difference 1.412 3 0.471 4.483 .007

MaterialAreaHeight difference 0.785 6 0.131 1.467 .196

Note:* df = degrees of freedom; MS = mean square; SS = sum squares.

NASA-TLX Assessment Performance Analysis

There was no significant difference in the time load or frustration. Table 3 shows items with significant differences.

Table 3.
All Significant Items in the Variable Analysis of Workload Assessment Performance.

Workload item Variable SS df MS F p

Intelligent load Operation zone area 41.763 1 41.763 18.341 .000

AreaHeight difference 32.518 3 10.839 6.827 .001

Physical load Operation zone area 22.816 1 22.816 11.020 .004

Height difference 15.912 3 5.304 3.603 .019

AreaHeight difference 26.886 3 8.962 5.997 .001

MaterialAreaHeight difference 29.877 6 4.980 2.278 .041

Mental load AreaHeight difference 9.386 3 3.129 4.068 .011

Self-performance Operation zone area 10.441 1 10.441 5.123 .036

AreaHeight difference 9.129 3 3.043 3.952 .013

Note: df = degrees of freedom; MS = mean square; SS = sum squares.

Intelligent Load

The intelligent load for large areas (2.4) was significantly better than that for small areas (3.1; p = .000). Interactions between the operation area ratio and height difference were present. When the area was large, models 1, 2, and 4 were significantly better than model 3 (p = .006 for model 1 vs. model 3, p = .039 model 2 vs. model 3, and p = .004 model 4 vs. model 3).

Physical Load

The physical load for large areas (2.3) was significantly better than that for small areas (2.7; p = .000). The height difference value of model 1 (2.2) was significantly better than that of model 3 (2.7; p = .014) and that of model 1 was also significantly better than that of model 4 (2.6; p = .012). Interactions between height difference and operation area ratio were present. When the area was large, models 1, 2, and 4 were significantly better than model 3 (p = .006 for model 1 vs. model 3, p = .025 for model 2 vs. model 3, and p = .005 for model 4 vs. model 3). Regarding the interactions among the three factors, the value for model 2 under the large area (1.8) was significantly better than that for under the small area (3.3) when FDM_TPE was used (p = .003). The value for model 4 under the large area (2.2) was significantly better than that for under the small area (2.7; p = .029). However, the value for model 3 under the large area (3.4) was not better than that for under the small area (2.2; p = .041).

Mental Load

Interactions were present between the operation area ratio and height difference. When the area was large, the values for models 1 and 4 were significantly better than those for model 3 (p = .034 for model 1 vs. model 3 and p = .043 for model 4 vs. model 3).

Self-Performance

The value for the large area was 6.2, which was better than that for the small area (5.9; p = .036). Interactions between the height difference and operation area ratio were present. When the area was large, the values for models 2 and 4 were significantly better than those for model 3 (p = .017 for model 2 vs. model 3 and p = .027 model 4 vs. model 3).

Discussion

The results of our experiment showed that lesser time was needed to identify large areas than small areas, and the operation area ratio would affect the identification performance. When the area was large, the identification time for models 1 and 4 was better than that for model 3, and the accuracy performance for models 1, 2, and 4 was better than that for model 3. The results were similar to those reported by Shimizu et al. (2000), who recommended not using hollowed plane surfaces. Regarding the NASA-TLX assessment scores, the intelligent load, physical load, and self-performance were better under the large area. The physical load results showed that model 1 was better than models 3 and 4. Under the large area, the intelligent load, physical load, mental load, and performance were the lowest for model 3.

When FDM_TPE was used for the physical load item, models 2 and 4 maintained good performance under the large area. However, model 3 showed the reverse results. As print tracks are more visible on touch in FDM_TPE graphics cards, FDM_TPE material may increase the number of steps in the identification of the pair of scissors. Therefore, the comparison of the aforementioned items found that a large operation area ratio may increase the identification performance. Among the models, model 3 is not recommended for large areas; model 1 had a superior physical load performance than model 4, and although models 1 and 2 (greater height) showed results similar to those reported by McCallum et al. (2006), who did not find any significant differences, the time spent for model 2 was longer. Therefore, model 1 is mainly recommended. Among the materials, FDM_TPE materials increased the number of steps in the identification of the pair of scissors and should not be used. As FDM platforms are inexpensive and easy to operate, it is recommended to use FDM_PLA over SLA_ABS.

Lmplications for Practitioners

A recommended procedure model for tactile-graphic design was obtained by combining the results of a previous study (Wu et al., 2021, in press) and this study (see Figure 3). The designer can first take a plan-view (perspective with the most complete operation zone shape) photograph of the object before using the computer graphics software to delineate the contours of the object, where a contour width of 1 mm is acceptable. After confirming the object size, a suitable presentation model can be used for adjustment (the identification performance is better if the operation area ratio of the object is large) before importing it to the 3D software. Following that, lines and planes can be shifted upward by 1 mm before conversion to a print file. FDM technology and PLA materials could then be used for printing.

Figure 3.
Procedural Model for Tactile-Graphic Design.

Limitations and Future Work

The tactile-graphic design procedure model proposed in this study was developed based on the examination of concrete household items. However, the impartation of knowledge is, at times abstract. For example, teaching students to identify the appearance of a pair of scissors is concrete but teaching students to use the pair of scissors is abstract. Teaching abstract concepts may be more difficult and should be further examined. With regard to the application of 3D printing, the greatest difficulty lies in how to enable teachers for the visually impaired to convert their designed graphics to printable files through existing software after the recommended procedure for creating graphics has been formulated. Perhaps simpler software for model software can be developed so that teachers can easily create 3D teaching materials.

Variable	SS	df	MS	F	p
Material	365.320	2	182.660	1.494	.238
Operation zone area	1414.055	1	1414.055	13.607	.002
Height difference	202.059	3	67.353	0.510	.677
Material*Area	663.741	2	331.871	1.649	.207
Material*Height difference	549.224	6	91.537	0.822	.555
Area*Height difference	960.989	3	320.330	3.688	.017
MaterialAreaHeight difference	451.083	6	75.181	0.723	.632

Variable	SS	df	MS	F	p
Material	0.083	2	0.042	0.562	.575
Operation zone area	0.219	1	0.219	1.670	.213
Height difference	0.474	3	0.158	1.616	.196
Material*Area	0.057	2	0.029	0.552	.581
Material*Height difference	0.303	6	0.050	0.685	.662
Area*Height difference	1.412	3	0.471	4.483	.007
MaterialAreaHeight difference	0.785	6	0.131	1.467	.196

Workload item	Variable	SS	df	MS	F	p
Intelligent load	Operation zone area	41.763	1	41.763	18.341	.000
Area*Height difference	32.518	3	10.839	6.827	.001
Physical load	Operation zone area	22.816	1	22.816	11.020	.004
Height difference	15.912	3	5.304	3.603	.019
Area*Height difference	26.886	3	8.962	5.997	.001
MaterialAreaHeight difference	29.877	6	4.980	2.278	.041
Mental load	Area*Height difference	9.386	3	3.129	4.068	.011
Self-performance	Operation zone area	10.441	1	10.441	5.123	.036
Area*Height difference	9.129	3	3.043	3.952	.013

Footnotes

Authors' Note

Hsiang-Ping Wu, Department of Industrial Design, Tatung University, Taipei; Department of Product Design, Ming Chuan University, Taiwan.

Chin-Te Chang, Taipei School for the Visually Impaired Teacher, Taiwan.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Ministry of Science and Technology (grant number NSTC 109-2221-E-236 -001).

ORCID iD

Hsiang-Ping Wu

References

Brulé

Bailly

Brock

Gentès

Jouffrais

. (2018). An exploratory study of the uses of a multisensory map—with visually impaired children. Multimodal Technologies and Interaction, 2(3), 36. https://doi.org/10.3390/mti2030036

Buehler

Comrie

Hofmann

McDonald

Hurst

. (2016). Investigating the implications of 3D printing in special education. ACM Transactions on Accessible Computing (TACCESS), 8(3), 1–28. https://doi.org/10.1145/2870640

Butler

Holloway

Reinders

Goncu

Marriott

. (2021). Technology developments in touch-based accessible graphics: a systematic review of research 2010-2020. In Proceedings of the 2021 Chi conference on human factors in computing systems (pp. 1–15). arXiv preprint arXiv:2102.01273. https://doi.org/10.1145/3411764.3445207

Carfagni

Furferi

Governi

Volpe

Tennirelli

(2012, October). Tactile representation of paintings: an early assessment of possible computer based strategies. In Euro-Mediterranean Conference (pp. 261–270). Springer. https://doi.org/10.1007/978-3-642-34234-9_26

Cavazos Quero

Bartolomé

J. I.

Cho

(2021). Accessible visual artworks for blind and visually impaired people: comparing a multimodal approach with tactile graphics. Electronics, 10(3), 297. https://doi.org/10.3390/electronics10030297

Cekic

Begic-Hajdarevic

Cohodar

Muhamedagic

Osmanlic

. (2019). Optimization of stereolithography and fused deposition modeling process parameters. Annals of DAAAM & Proceedings, 30, 681–687. https://doi.org/10.2507/30th.daaam.proceedings.093

Elkhuizen

W. S.

Callewaert

T. W.

Leonhardt

Vandivere

Song

Pont

S. C.

Geraedts

J. M. P.

Dik

. (2019). Comparison of three 3D scanning techniques for paintings, as applied to Vermeer’s ‘Girl with a Pearl Earring’. Heritage Science, 7(1), 1–22. https://doi.org/10.1186/s40494-019-0331-5

Ferracane

M. F.

(2020). Redesigning Traditional Education. In Redesigning Organizations (pp. 329–343). Springer. https://doi.org/10.1007/978-3-030-27957-8_25

Gual

Puyuelo

Lloveras

. (2015). The effect of volumetric (3D) tactile symbols within inclusive tactile maps. Applied Ergonomics, 48, 1–10. https://doi.org/10.1016/j.apergo.2014.10.018

10.

Hart

S. G.

Staveland

L. E.

(1988). Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research. In Advances in psychology (Vol. 52, pp. 139–183), North-Holland. https://doi.org/10.1016/S0166-4115(08)62386-9

11.

Holbrook

M. C.

Andonova

(2006). GRASP: Graphic research and standards project. Proceedings from the International Council for People with Visual Impairments Conference.

12.

Holloway

Marriott

Butler

(2018, April). Accessible maps for the blind: Comparing 3D printed models with tactile graphics. In Proceedings of the 2018 chi conference on human factors in computing systems (pp. 1–13). https://doi.org/10.1145/3173574.3173772

13.

Jehoel

Sowden

P. T.

Ungar

Sterr

. (2009). Tactile elevation perception in blind and sighted participants and its implications for tactile map creation. Human Factors, 51(2), 208–223. https://doi.org/10.1177/0018720809334918

14.

Mardis

N. J

. (2018). Emerging technology and applications of 3D printing in the medical field. Missouri Medicine, 115(4), 368.

15.

McCallum

Ungar

Jehoel

. (2006). An evaluation of tactile directional symbols. British Journal of Visual Impairment, 24(2), 83–92. https://doi.org/10.1177/0264619606063406

16.

Panjatevakupt

(2020). Your tactile story tray: collaborative tactile images to share with blind and visually impaired people. [MRP]. OCAD University Open Research Repository.

17.

Shimizu

Shinohara

Nagaoka

. (2000). Recognition of tactile patterns in a graphic display: Evaluation of presenting modes. Journal of Visual Impairment & Blindness, 94(7), 456–461. https://doi.org/10.1177/0145482X0009400706

18.

Stangl

Hsu

C. L.

Yeh

(2015, October). Transcribing across the senses: community efforts to create 3D printable accessible tactile pictures for young children with visual impairments. In Proceedings of the 17th International ACM SIGACCESS Conference on Computers & Accessibility (pp. 127–137). https://doi.org/10.1145/2700648.2809854

19.

Urbas

Pivar

Elesini

U. S

. (2016). Development of tactile floor plan for the blind and the visually impaired by 3D printing technique. Journal of Graphic Engineering and Design, 7(1), 19–26. https://doi.org/10.24867/JGED-2016-1-019

20.

Wonjin

Jang

H. I.

Harianto

R. A.

J. H.

Lee

H. J.

Moon

(2016). Introduction of 3D printing technology in the classroom for visually impaired students. Journal of Visual Impairment & Blindness, 110(2), 115–121. https://doi.org/10.1177/0145482X1611000205

21.

C. F.

H. P.

Y. H.

Yeh

I. T.

Chang

C. T.

(2022). Constituent elements affecting the recognition of tactile graphics. Journal of Visual Impairment & Blindness, 116(2), 194–203. https://doi.org/10.1177/0145482X221092031