Creating 3D picture books for visually impaired and blind persons
Ben Armstrong, Brett Edwards, Nathan Flynn, Hannah Fjellman, and George Halunen
Abstract
The team developed a process to translate 2D children’s books into 3D versions that recreate the drawn images into tactile images for the purpose of making 2D children’s books accessible to visually impared children. The team created 3D images of the pages, translated the writing into braille, and then printed the pages using PLA. Ultimaker S3 and MiniLab printers using FFF extrusion were used to print the pages. The team recommends using Ultimaker S3 printers in the future because of their great print quality. While there is still significant room for improvement, the process outlined in this paper is a good start to making children’s books more accessible to visually impared children.
Introduction
Children’s books in braille can be difficult to find. Accommodations typically just include braille, not updated images of existing books for visually impaired children to enjoy. Our team’s goal for this project was to develop a process that could transform pre-existing books into 3D versions with braille lettering, and to use this method to re-create existing books for visually impaired children.
Materials
PLA was selected as the material to manufacture the 3D printed pages. This material was selected due to its inexpensive printing price, which is beneficial for the large volume of printing the group needed to perform to complete the full book, as well as the fact that this material was readily available in both the Makerspace lab as well as the Polymer Engineering Center lab. The group considered using Nylon as the printing material as well. This is due to the fact that Nylon is more flexible and would more closely resemble a turnable book page. Nylon was not used in any testing during the scope of this project due to the time constraint of printing the full book in the PLA material.
Page Design
Braille Integration
The team decided that transcribing the pages from English to braille would be the best fit for having words on the 3D printed pages. Rather than a language, braille is almost a 1:1 transcription of English. Each English letter can translate to one braille character. Of the six dots used in a braille character, some of the dots are raised to denote a letter, digit, or punctuation mark. See Figure 1 for the standard English Braille Alphabet, also known as Grade 1 braille.
Figure 1: Standard Grade 1 Braille
There are three different “Grades” of braille, all differing slightly. Since Grade 2 is most commonly used in the US, the team decided to use Grade 2 Braille on all of the pages. In addition to the 26 English letters, 10 digits, and common punctuation, Grade 2 Braille also includes contraction characters. These contractions make Grade 2 Braille words shorter than Grade 1 Braille words, which do not include contractions. See Figure 2 for the contractions used in Grade 2 Braille. Grade 3 Braille is the least commonly used and is not taught in schools, so it was not considered for this project.
Figure 2: Standard Grade 2 Braille Contractions
To transcribe the English sentences in “Sunflower Lion”, a Java program was written that outputs a PNG file containing the braille for the corresponding page. See Figure 3 for an example. The program correctly spaces the braille characters and dots within each character according to the English standard.
Figure 3: Example of Braille PNG Output by Java Program
SVG File Creation
To translate the 2D book pages into a 3D model, a repeatable procedure was developed. A simplified model of the procedure can be seen in Figure 4.
Figure 4: Page modeling model
First, the 2D book pages were scanned into a PDF format. The book pages were then brought into Word and centered on the page. The Word pages used a square format with a black border around the page. Next, a cropped version of the braille PNG file for that page was pasted over the corresponding English words. An example of the page creation process is shown in Figure 5.
Figure 5: Example of completed page in Word
Next, that page was converted into a PNG. The PNG was then converted into an SVG file. SVG file conversion works by tracking lines around shapes and then turning them into profiles. The file conversion procedure is shown in Figure 6.
Figure 6: SVG to 3D file
When the PNG file was converted to SVG a black border was needed to define the outside of the page. To convert the file to a 3D file, an open-source free cad software, FreeCAD, was used. Using a SVG import feature in FreeCAD, we were able to import the SVG file and have it convert all of the profiles into objects that could be extruded. The braille characters were set to a height of 1 mm, and the other shapes on the page were set at heights up to 4 mm. Once all of the important shapes and characters were extruded, the STL file was exported and ready to 3D print.
Fabrication
Ultimaker S3 printers at the Makerspace were used to print some of the pages at an 8 in by 8 in. size. Uniform scaling was turned off in Cura to leave the Z height of the file alone and adjust the X and Y size to enable the largest print possible. Pages were then printed one by one. Each team member was allowed to reserve only 2 printers concurrently due to the Makerspace’s high demand. Each print using the Ultimaker took between 5 and 6 hours to complete. An example of the printed page printed in white PLA can be seen in Figure 7.
Figure 7: Test print done at the MakerSpace
The team used four MiniLab printers in the polymer lab. Each printer had a different nozzle diameter which were 0.4 mm, 0.5 mm, 0.6 mm and 0.8 mm. A different slicing program was used to prepare the STL file, but the files were prepared using the same general process. Using the polymer lab, multiple pages could be printed at one time. An image of a page printed on the polymer lab machines is seen in Figure 8.
Figure 8: A test print done at the polymer lab
Results
The pages printed in the Makerspace printed with a 0.4mm nozzle resulted in a high resolution image with smooth surface finish, seen in Figure 7. The braille dots were also a high enough resolution to be distinguishable when touching with a finger and did not display any defects. The downside of the slow print at small nozzle diameter was that each print took close to 6 hours to complete. With only 2 prints being allowed at one time and over 20 pages to print, this meant that printing the pages with the Ultimakers took a significant amount of time.
The pages from the polymer lab varied in surface finish depending on the nozzle diameters. A benefit of these printers was the faster movement speed of the nozzle head with significantly reduced print time. These prints averaged around 90 minutes per page, and the group was able to utilize 4 printers simultaneously. A negative outcome from some of the prints with larger nozzle diameters was the braille dots being connected with strings of filament as seen in Figure 9.
Figure 9: Failed print with disfigured braille dots
These prints would be unusable for a braille user as there is no tactile difference when feeling along each line. Using a slower print speed or smaller nozzle diameter was a solution to eliminate this defect.
Future Recommendations
In the future, it is recommended that a different printing process be used. FFF does not have the resolution for the finely detailed braille. Also, FFF does not create the smooth surface texture that is desired. Other potential processes are selective laser sintering (SLS) and stereolithography (SLA). SLS can print multiple pages at one time by stacking them on top of each other. This would reduce the print time and setup of SLS. SLS has a higher resolution than FFF which would allow the braille to be printed more clearly. SLA would create a smoother surface finish than FFF but it would be difficult to fit a page on the printing bed. Along with SLS, SLA has a finer resolution than FFF. Although, if FFF were to be used, it is recommended a smaller nozzle than 0.4 mm be used for finer detail.
The group recommends creating an automated system for elevating the picture elements. The current process is manual as elements have to be manually extruded in FreeCAD. Also, PDF images of the books have to be first converted to a PNG and then an SVG. An automated system would skip all the manual editing of an image. One possibility is creating a script that prepares the image for CAD. Ideally, the system could prepare the file so when it is opened in CAD the page would already be elevated. Another option could be using machine learning to analyze all the book pages and decided what needs to be elevated. These two options are proposals to create an automated system for elevating the pictures. If the manual process must be used, then acquiring PDFs of the book pages is recommended.
A new material besides PLA would be another recommendation. The PLA used was breaking because of the thin pages, thus trying a stronger material would be advantageous. A flexible material, similar to TPU, would mimic the hand feel of a traditional book. The user would be able to turn the pages with a flexible book rather than flip them with static PLA pages. Flexible material would overall enhance the experience of a blind or visually impaired person. Testing with various materials would be a necessary future step.
Finally, the group recommends an Arduino be integrated into the book. The setup could include a button on each page and a voice playback module. Once a button is pressed the playback would read the page. This would allow the child and their guardian to understand the story, regardless of their braille literacy. The Arduino could be hidden on the binding of the book or in the back of the book.
Conclusion
The group successfully completed a full 3D printed book using FFF printers. A manual process of taking a children’s illustrated book and converting it into usable CAD files for use in a 3D printer was created, but there is much room to improve. An automated process would significantly reduce manual time spent on this step and machine learning could allow for all kinds of children’s books to be converted. Converting sentences to braille was successful but would also benefit from an automated process due to the long processing time. Prints using the Ultimaker S3 printers achieved a high resolution but at the cost of a 6-hour print time. Prints done using the Minilab printers with larger nozzle diameters also achieved acceptable surface resolutions, but the fastest prints did come at a cost of defective braille dots. There is much room to improve outside the scope of this project, but the market impact and the human impact is great and will allow for another avenue of learning and experience for children in the visually impaired community.
References
“What is braille? [your guide to braille],” Braille Works, 17-Jan-2022. [Online]. Available: https://brailleworks.com/braille-resources/what-is-braille/. [Accessed: 16-Feb-2022].