Exploring the design of a tactile reader for braille and non-braille users

Abstract

BACKGROUND:

Research focusses on providing economical and effective tactile readers which incorporate relative motion between the fingertip and the reading text.

OBJECTIVE:

Here, a novel design and prototype of a tactile reader is proposed for blind Braille and non-Braille users based on the sequential reading of text when using a single cell reader.

METHODS:

The accuracy of reading words using this design was evaluated on 17 visually challenged users who know Braille. With a slight modification in the design to benefit non-braille users, the reading of words on 10 blind-folded sighted individuals and 12 late-blind visually challenged users was investigated. Differences in recognition of English letters formed by dots and lines were explored.

RESULTS:

The results showed no significant difference in accuracy when reading words with this prototype compared to braille reading on paper. The percentage accuracy in reading words for the 39 users was 98.62% for blind braille readers, 90% for blindfolded sighted users and 96.8% for blind non-braille users.

CONCLUSION:

The findings indicate that the proposed design for tactile reading can potentially benefit both groups of visually challenged users-those who know braille and those who do not.

Keywords

Braille users non-braille users laterotactile reading

1. Introduction

Braille is a language of the blind invented by Louis Braille in 1825. A Braille character consisted of 6 dots, arranged in 2 columns of 3 rows. Every letter is depicted by patterns of raised and lowered dots. A visually impaired user reads by scanning the patterns of raised and lowered dots. When accessing information from the computer, a large number of visually impaired people mainly use speech software. Yet since reading text is quite different from listening, researchers are looking to develop refreshable braille displays which electronically raise and lower the dots of braille cells when representing information from the computer. The widespread use of refreshable braille displays has been hindered by its cost and much research has gone into making refreshable braille displays cost effective which can be accessible to everyone in society. The most common commercially available refreshable braille display uses piezoelectric actuators [1, 2, 3]. The cost of these braille displays ranges from $500 to $15000 based on the number of cells being displayed. A display with fewer braille cells will cost less as fewer number of actuators are needed to position the dots. Recently produced refreshable braille displays such as the Canute [4] provide 9 lines of forty characters of Braille but details of user testing and feedback are not available. A recent review [1] categorizes various designs of tactile readers based on the types of actuators used such as piezoelectric actuators, Shape Memory Alloy actuators, Thermopneumatic and Electroactive polymer actuators.

Experienced braille readers use both hands to read Braille on paper using the butterfly technique, where both forefingers read together for approximately 2/3 of a line, then the right-hand forefinger reads ahead to the end of a line while the left-hand forefinger runs back to begin reading the next line. This technique is said to be the most efficient method of reading braille [5]. However, this method of reading will only be supported on a full-page braille display and not on a single line refreshable braille display or using a single-cell device. Even so, there is still a need for early braille learners and mid-level braille learners to have access to affordable tactile readers [6] and therefore research continues to make tactile readers effective and affordable.

Alternate ways of displaying single cell-braille, such as the Ubibraille and the V-braille have also been researched [7] but designs involving alternate modes of display are not commercially available as yet. In its standard form, braille is represented in 2 columns of 3 dots each, with standard spacing specifications as mentioned in [8]. In these alternate methods of displaying braille, the 6-dot format is not adhered to in terms of the arrangement of the dots naturally forcing the user to assimilate braille in a different way. These alternate displays also present a vibratory stimulus to the skin of the user instead of sliding traction or the normal indentation as seen in other designs.

Refreshable braille displays vary in cost depending on the number of braille cells depicted as the number of actuators proportionally increases. Decreasing the price of a braille display can be done by reducing the number of actuators, limiting the display to a single cell and refreshing it according to the text it depicts. While the speed of reading with single cell readers decreases, the trade-off is a decrease in the cost. In using single cell readers, braille readers construct the word sequentially, processing the word character by character – as their tactual input is limited to their primary reading finger. Designs of this type, such as the tactile mice, have been produced [9], but have not been proven to be effective in reading text as there is no lateral movement of the finger pad over the braille pins as seen in regular braille reading [2, 10]. Similarly, in the recent design of a single cell tactile reader, the BraiBook [11] there is no sliding movement between the finger pad and the text. Foulke and his team carried out extensive studies on braille reading and found out that pressing letters into a stationary fingertip produced little deformation compared to normal modes of braille reading. So, another instrument was developed that presented braille symbols on a tape. This was used by Kilpatick (cited in [2]) who concluded that an increase in performance was obtained when there was lateral motion between the fingertip and the braille reading surface. These results were further corroborated recently by Russomanno et al. [2] who, in their experiments, suggested that braille displays incorporating sliding movement between the reading finger and the braille text were more effective than having a braille display which did not allow this sliding contact. They also showed experimentally that proprioceptive cues obtained when using the hands enhanced the capacity of the user to identify letters.

Earlier laterotactile sliding displays included tape displays such as a bubbled tape loop [12] into which protrusions could be made by actuators and felt by the user, yet details on testing are not available. Another inspiring work on rotating wheel displays was seen in [13]. This wheel display used 3 actuators for 6 dot braille and 4 actuators for 8 dot braille display. The pins were set before they came under the area of the users reading finger at a constant speed. However, reversals were not accommodated and user assessment on the device was not fully investigated. The NIST Braille display is a further extension of this work. It has a rotating drum through which the pins are actuated and depict the specific braille character. The design here includes 400 pins with several rotating parts. Passive pin retention disks are used here [14].

A novel idea of giving the user the experience of the finger sliding over a braille dot was seen using the STRESS2 tactile transducer [15, 16] where progressive waves of skin deformation were caused by using deflecting piezoelectric benders. However, on testing with users, character recognition was stated to be low with an average of 57%. In another design, rotating wheels with a pair of rotating actuators [17] were used. It had portions of embossed braille characters on it so that when the pair was read together, entire braille characters could be read but there are no details on the assessment of this device.

Figure 1.

Experimental set up [23].

Another factor that must be taken into consideration in the design of tactile readers is the decrease in the number of visually impaired users who actually know braille [18]. A large percentage of those who are blind do not actually know braille. Braille is generally learnt at an early stage while it takes longer for those who are late-blind to learn braille [19]. As early as 1933, it was recognized that late-blind retain visual imagery. Therefore, a font using Moon type characters were developed which were halfway between braille and English letters with some arbitrary characters and could be used by the late-blind. The Moon type is the only system besides Braille which existed for a substantially longer time than any other system of tactual reading [20]. Earlier studies as in [21] compared the tangibility of letters and braille and stated that braille dots were more perceptible than letters. Yet similar studies by Heller et al. [22], where braille, letters and the Morse code were compared, found that with unrestricted exploration, and letter sizes of 8–11 mm, there was similar tangibility between braille and letters. This was perhaps due to the fact that exploration of letters was restricted in [21] and the size of the letters were small.

In this paper, a novel design is proposed, using just a single actuator and also allowing for sliding movement between the text and the finger pad. If words can be presented and perceived accurately with a single actuator, such designs would be economical and effective. A modification to the prototype is also presented that allows non braille users to access it. The results presented here provide information to guide the design of effective, economical, refreshable tactile displays that can be used by visually-impaired users who know braille and those who do not.

2. Description of the prototype

The prototype designed and developed consisted of two parts-an embossed disc that was mounted on the shaft of a stepper motor and the interfacing driving electronics within an enclosure. The experimental set-up is shown in Fig. 1. The enclosure holds the driving electronics and bears the weight of the hand resting on it while the reading finger is placed on the disc. The reading finger is stationary while the tactile display rotates underneath it, similar to the placement of our hand over a mouse while interacting with the PC. Generally, in using a mouse, the users right index finger rests lightly on the left mouse button. In the design adopted here, it rests on the disc.

2.1 Control system

The prototype uses the TM4C123GH6PM microcontroller which is ARM Cortex-M4F based processor core. The composite USB port includes a UART connected as a virtual serial port by which serial communication can take place between the computer and the prototype. A booster pack was designed to fit the launch pad vertically. The interfacing electronics receives power supply from a 12 V, 2 A SMPS adapter to drive the stepper motor according to the pulses received via the microcontroller. The stepper motor is rated at 12 V with a step angle of 1.8 degrees and a holding torque of 0.303 N-m and weighs 220 g. It has a speed of 150 rpm. The control of the motor here is open-loop and is programmed to run in unipolar, full stepping sequence. At this speed, 400 ms is the maximum time taken to display character separated by a full 360 ${}^{\circ}$ .

Characters are transmitted from the computer at a baud rate of 115200 using the freeware application Hercules for serial port communication. As the characters are sent out by the experimenter via the serial port, the motor rotates bringing the required character underneath the users reading finger. The user reads the characters by touch.

Figure 2.

Embossed braille disc (dimensions in mm).

2.2 Braille embossed tactile display

The embossed disc is 3D printed and made of ABS (Acrylonitrile Butadiene Styrene) plastic which is an opaque thermoplastic and amorphous polymer. Initially, 25 braille characters were embossed around its circumference with a braille character placed every 14.4 ${}^{\circ}$ . As preliminary testing with the visually challenged users showed that the braille characters were too close to each other, the disc was modified with only alternate braille characters embossed i.e. a braille character was embossed every 28.8 ${}^{\circ}$ (16 steps). This resulted in 12 braille characters being presented to the user. With an arc length, s, of 5 mm assigned to each braille cell and the angle between the characters taken as $\Theta=$ 28.8 ${}^{\circ}$ , the radius of the disc calculated is 20 mm. The disc has a width of 10 mm as in Fig. 2, which is sufficiently wide enough for users to place their reading finger on. The disc is mounted on the motor shaft and rotates by different angles according to the letter pressed on the keyboard of the laptop. The disc is placed on the motor shaft which has a diameter of 5 mm. The height of the braille dots is 0.9 mm and the width is 1.4 mm in accordance with the standard values [8]. It has a fillet of 0.4 mm to give it a curved smooth surface similar to that of braille dots. The 12 braille characters embossed on the disc are o, l, h, r, c, m, n, k, d, g, n, and “ ”. These characters were chosen at random and could form simple words in a variety of combinations. Twelve characters were used for a start, to see if this method of reading was viable. The ‘space’ is also a character, as a space distinguishes one word from another. Twenty-four words (2-letter, 3-letter, 4-letter, 5-letter words) can be formed using these letters. The outcome of testing with these 12 braille characters is documented in the results. Appendix has a list of words used for testing with this braille disc. The results of preliminary testing with this prototype using the braille embossed disc is seen in [23].

2.3 English letter embossed tactile display

The braille embossed disc was replaced with English letter embossed discs which are described below as an adaptation for non-Braille users. The 12 English letters embossed on both the discs were I, T, C, H, O, E, S, K, N, A, D. These characters were chosen as simple words could be formed with them. It was of interest to find out if letters formed of lines or dots were more perceptible to the user in this method of reading words. Hence, two types of 3D printed English letter embossed discs were used-Lined and Dotted discs (Figs 3 and 4). Similar to the design of the braille disc, a space character distinguished one word from another. The Calibri, Size 34 was the font type used to emboss lined letters on the disc. Each letter was extruded by a height of 1 mm. Uppercase letters were embossed as prior research [21] highlighted increased tangibility of such letters in comparison to lowercase letters.

Figure 3.

Embossed lined letter disc (dimensions in mm).

Figure 4.

Embossed dotted letter disc.

Previous research has shown that dots are understood to be more tangible than lines [20]. Therefore, for the dotted disc, since there was no available font in SolidWorks, 12 dotted letters were drawn, dot by dot, as seen in Fig. 4. The dot diameter was 1.4 mm, extruded by 1 mm (a requirement of the 3D printer) with a fillet of 0.4 mm for a curved surface. The spacing between the dots was kept at 2 mm generally. Yet in some letters such as ‘S’, the spacing between the dots was 1.4 mm. The angle between letters was 28.8 ${}^{\circ}$ similar to that of the braille embossed disc. The 12 English letters embossed also included the space character, as words can be recognized as groups of letters separated by spaces. Appendix has a list of words used for testing with the English lettered disc.

In this design the sliding friction provided between the rotating disc and the finger-pad contributes to the perception of the character. The design allows for both a tangential traction and normal traction of the finger-pad as the character on the rotating disc slides in underneath the finger-pad. Both the ‘sliding in’ of the character and the ‘resting under’ the finger-pad contribute to the perception of the character in this design. While the sliding in of the letter under the finger-pad is for a very short time and is dependent on the stepper motor characteristics, the time that the letter rests underneath the finger-pad depends on the user. The direction of rotation of the disc forces the finger to brush over text just as in conventional braille reading. This design also allowed for vertical movement of the finger pad if needed. It was observed that letters such as ‘K’ or ‘N’ needed vertical exploration of the finger-pad too while letters such as I, T and O were perceived easily. With the method of reading involved here, only the letters that contribute to the word stop under the primary reading finger. If the next letter is on another portion of the disc, then the user touches all the embossed letters in between, till the disc stops its rotation at the specified letter. However, since the rotation of the disc is fast, these in-between letters are not perceived. Since the motor is 1.8 degrees per step which is 200 steps per revolution (2 ms per step), the longest delay between the appearances of the same character under the reading finger would be 400 ms, equivalent to 2.5 Hz. In the disc design here, the smallest number of steps between consecutive letters of a word is 16 steps (since the letters are embossed every 16 steps), equivalent to 32 ms or approximately 32 Hz. Therefore, the refresh rates of letters contributing to the word vary from 2.5 Hz to 32 Hz here. Prior work states that 1 Hz would be just enough for slow reading although 10 Hz is recommended for most applications [8]. The refresh rates using this method of reading vary between 2.5 Hz and 32 Hz. The weight of the hand rests on the enclosure and therefore the finger lightly rests on the surface of the disc and does not impede the motion of the rotating disc. Earlier studies have shown finger force to be 5–10 g [21]. It must be noted that the placement of the reading finger is similar to how one places the hand on the computer mouse. The design of the prototype was such that it allowed horizontal and vertical exploratory movement of the primary reading finger, if needed, for identification of the letter. In effect, while it passively received the characters, an active exploration of the letter was allowed. Consider the word H-O-O-K in the English letter disc as seen in Fig. 5. The time in milliseconds between the appearances of letters under the reading finger are shown. It can be seen that the longest delay of 400 ms is between the appearance of the letter O since the motor needs to rotate a full 360 degree for ‘O’ to come back under the reading finger. It should be noted that between the consecutive appearances of letter ‘O’, 11 nonsensical characters pass under the reading finger. Whether this aspect of the design prototype causes a hindrance in reading is explored in the experiments that ensue.

Figure 5.

Time in milliseconds between the appearances of letters.

3. Methods

3.1 Testing on blind braille (BB) users

3.1.1 Participants

Seventeen (4 F and 13 M) blind braille readers, (mean age 21 years, range 12–49), participated in the experiment. Fifteen of them were fully blind while the rest had low vision since birth. Five were left-handed and used their left index finger to read while the rest used their right index finger as their primary reading finger. All of them reported to spend an average of 2 hours per day reading Braille apart from the time they spent at school/work place. Nine readers read paper braille at speeds greater than 50 wpm and the remaining 8 read braille at less than 50 wpm. Nine of them read paper braille with both hands yet adapted to reading with the prototype easily possibly because of the use of their primary index finger when reading. The medium of instruction for Braille at their school/workplace was English.

3.1.2 Procedure

Braille readers are accustomed to reading by touch and therefore, required a short time to get used to the feel of the prototype and the placement of their primary reading finger on the embossed disc. As part of a training session, they were allowed to recognize letters and words as the experimenter initially sent out letters and words via uart, so that they would get accustomed to reading in this way. The prototype was connected to a laptop during testing and the participants placed their hand on the device with their primary reading finger resting lightly on the disc as shown in Fig. 1. The presentation of characters was via hyper terminal sent by the experimenter. For the testing session, the readers were asked to read words from the experimental set-up in the same way that they read from paper braille. If a character could not be understood, they were instructed to say ‘Pass’ so that the next letter could be sent via uart. Three sets of 24 randomly assorted words in all (2-letter, 3-letter, 4-letter, 5-letter words formed from the 12 characters) were presented to the user as Trial1, Trial2 and Trial3 in two sessions, Session1 and Session 2. A short gap, ranging from a few hours to an entire day, between the 2 sessions was given to participants. The accuracy of reading out words was calculated as the number of correct responses divided by the total number of words read within a time frame of 1 minute. This was similar to the test of braille reading employed in [24] where they recorded the number of words read out in a minute. The mean percentage accuracy over the three trials was then calculated. The user’s accuracy on a printed braille sheet which had similar words was assessed. The accuracy on both paper braille and the prototype was then compared within the same time frame of 1 minute. The speed of the session was noted too.

3.2 Testing on blind folded sighted (BFS) users and Blind Non-Braille (BNB) users

The prototype was slightly modified to include testing with users who did not know Braille. The braille embossed disc was replaced with the English letter embossed disc. The experimental procedure which is described below was used for both BFS and BNB users. With this experiment, the tactual reading of words for those who do not generally read by touch is evaluated. A common feature between the blind-folded sighted subjects and the late blind users is that they both retain a visual memory of the alphabet. Twelve BNB users and 10 BFS users volunteered for the study and results of the testing are documented here.

3.2.1 BFS participants

Ten blind-folded sighted subjects (6 F and 4 M) volunteered for the study (mean age 20.7 years, range 14–43). All of them were right-handed and used their right index finger as their primary reading finger, similar to placement of their hand on the mouse of a PC. While two of the participants were from High school, the rest were employed/students at university. All the participants were proficient in English. All the participants were accustomed to using a mouse and so were comfortable with the experimental set-up. Even though they have an active memory of the shapes of letters, they are not accustomed to reading by touch and required training sessions to familiarize themselves with reading tactually. Therefore, in testing with blind-folded sighted users, in addition to documenting the accuracy, the number of training sessions it took each participant to obtain 90% accuracy in reading English characters was also noted.

3.2.2 BNB participants

Twelve male subjects volunteered for the study (mean age 15 years, range 12–21). All of them were right-handed and used their right index finger as their primary reading finger in this study. All of them were partially-blind and spent an average of 3.4 years in a sighted school before joining a blind school, except 1 university level student who had his friends/family read text out to him. Four of these participants used magnifiers to read hard text, two used only screen readers, one used both magnifier and screen reader and five volunteers used neither. The medium of instruction was English. Nine of the volunteers were very proficient in English and so were tested on words of character length 2, 3, 4 and 5. The three less proficient ones were tested on words up to 4 characters in length. They had a residual memory of the shapes of letters as they had spent a few initial years in the sighted school. Also, their interaction with English via screen reader/magnifier kept the afresh the memory of the shapes of letters.

3.2.3 Procedure

BFS users and BNB users are not accustomed to reading by touch. They required training sessions to familiarize themselves with recognizing letters by touch augmented by the memory of the shapes of letters. The training session consisted of 25 letters (jumbled 12 embossed letters) being randomly sent out via uart at timed intervals of 5 s and 3 s via a Windows application written in C#. The number of sessions it took to obtain 90% accuracy was noted. Every training session lasted less than thirty minutes. This was done for both the discs-Lined Letter disc and the Dotted letter disc. Once 90% accuracy was reached in recognizing letters, the training session concluded. The next level of testing was with words. The number of training sessions needed and the percentage accuracy in reading words was recorded.

The words were formed from the 12 embossed letters as seen in the Appendix. During testing the experimenter sent out letters that constitute the word using the hyper terminal. If the user didn’t recognize a letter, they were instructed to say ‘Pass’. The number of correct words read out loudly in a minute were noted. The speed of the reading session of the user were noted. Accuracy was calculated as the number of correct responses read out loudly by the user divided by the total number of words sent out by the experimenter in a time frame of 1 minute. The percentage accuracy was then calculated. This procedure was done using both types of discs. The same procedure was used for the study with 12 Blind Non-Braille (BNB) users except that the 9 proficient English users were tested on 2, 3, 4, 5 lettered words while the remaining 3 were tested on 2, 3, 4 lettered words.

3.3 Extended study on a single BFS user

This study was done on a single blindfolded sighted volunteer JT (Male, Age 15 years) due to the difficulty in finding a volunteer from the blind community over an extended period of time. JT was right-handed and used his right index finger as the primary reading finger here, placing it on the disc. He was proficient in English being a high school student and was adept at using the PC with both mouse/key control. JT had previously volunteered in the cross-sectional study of BFS users described in Section 3.2. His session speed reported using the dotted disc was 7 wpm and on the lined disc was 5 wpm. His accuracy using both types of discs were 100%. Even though his session speed was higher using the dotted disc, JT preferred using the lined disc for the extended study.

3.3.1 Objective of this study

In this extended study, two questions were explored-whether there was an increase in speed when the user had a full control of the reading session and whether there was a progressive desensitization of the fingertip when it was placed on the rotating disc, as indicated by worsening accuracy.

Table 1
Comparative results in percentage word accuracy for 17 blind-braille (BB) users

Mean percentage word	SD	Mean percentage word	SD	Mean percentage word	SD
accuracy paper braille		accuracy braille disc-Session1		accuracy braille disc-Session 2
96.152	4.03	97.21	6.02	98.62	3.84

Figure 6.

Testing procedure-percentage accuracy and speed tested in segment S1–S5.

Figure 7.

Percentage accuracy of 17 blind braille (BB) users.

3.3.2 Testing procedure

Ten test sessions (TS1–TS10) were conducted, two to three days apart depending on the convenience of the volunteer. Every test session was 20 minutes in duration. A timer maintained the duration of each of the segments. Each test session consisted of 5 segments S1–S5 of one-minute duration each as seen in Fig. 6. During each segment S1–S5, the display of letters was by a Windows App under user control via the arrow keys. The speed in words per minute and the corresponding percentage accuracy during these one-minute segments was noted. In between segments S1–S5 of each test session, the user was not in control and the disc kept rotating under the reading finger via the Windows App. JT switched between segments via keypress on the Windows App. JT was instructed to keep his finger over the rotating disc for the entire time of the test session in the same way as when he read during the Segments S1–S5. The backward arrow key was used in case reversals were needed to reconfirm any letter during the reading segments S1–S5 if needed. In the testing procedure, 2, 3, 4, 5, 6 lettered words were used. The Appendix lists the words used in testing. The words were read out loud by JT as in the previous testing procedures. Dividing the test session into segments helped to analyze the trend within each test session with respect to accuracy in reading. This was to ascertain if there was a desensitization produced in the finger over time by the rotating disc as indicated by worsening accuracy over different segments of testing. It also kept up the interest of JT over the 20-minute session as every 5 minutes there was a speed and accuracy check.

Table 2
Comparative results using both disc types for 10 BFS users

Disc type	Number of sessions to attain	Mean session	SD	Mean percentage	SD
	90% character recognition	speed WPM		word accuracy
Dotted characters	1–2	6.7	1.05	90	13.96
Lined characters	1–2	7	0.94	90.04	9.772

Figure 8.

Percentage accuracy of 12 blind nonbraille (BNB) users.

4. Results

4.1 Results-BB users

The accuracy of reading words for 17 users on paper braille and the experimental set up were high at an average of over 96% as seen in Fig. 7 and Table 1.

A statistical analysis of the results using a $T$ -test between Accuracy in Session 2 and on paper braille showed no significant difference ( $p=$ 0.092) indicating similar accuracy in reading words using paper braille and the prototype. The Bayes factor for a paired one sample $T$ -test computed using a web program [25] was in favour of the null hypotheses further corroborating the results. There was no significant difference in reading accuracy between Session 1 and Session 2 also. As seen in [26], single cell reading speeds for slow readers is stated to be 12 wpm. As stated previously, here the average session speed of 9.13 wpm is not the users speed of reading as time is shared between the experimenter and the user during testing. Yet we assume speeds will increase when the user is given full control of the session and they get more practice reading this way. This aspect is looked into in the extended study described further on. The users were asked to rate the reading method using this prototype as difficult or easy to use. All the users reported the ease of use in reading this way. None of the users mentioned any distraction caused by the brushing of nonsensical characters under the reading finger. It is likely it is due to the short delays between the appearance of the letters.

4.2 Results-BFS users

The results in Table 2 show an average accuracy of 90% with both types of English Lettered discs at 6–7 wpm after just 1–2 training sessions. These results are encouraging in comparison to the results obtained in [23] where blindfolded sighted users after 9 months of training read braille at the rate of 6–7 WPM. These results suggest that the training time for the tactual reading of non-braille users can be reduced if they are given raised print instead of Braille. It is reasonable to infer that the memory of the shapes of the letters have contributed to the reduction in training time needed to read tactually. A statistical analysis of the results using a $t$ -test showed no significant difference between the accuracy in reading with the dotted disc and the lined disc ( $p=$ 0.496). This result was further confirmed using the Bayes factor analysis. Similarly, a Bayes factor analysis and a $t$ -test done showed no significant difference ( $p=$ 0.196) in the speed of reading using the different discs. Therefore, when using this design, the disc can be designed with lettered or dotted characters.

Table 3
Comparative results using both disc types for 12 BNB users

Disc type	Number of sessions to attain	Mean session	SD	Mean percentage	SD
	90%-character recognition	speed WPM		word accuracy
Dotted characters	1	6.7	1.05	96.6	6.8
Lined characters	1	7	0.94	97.1	8.7

Figure 9.

Average speed during each test sessions TS.

4.3 Results-BNB users

A high accuracy in reading words is seen for the 12 BNB using both types of discs as in Fig. 8 within the same time frame of 1 minute. While the mean percentage accuracy of reading with the lined disc is higher than that of the dotted disc, a paired $t$ -test done on the two readings shows no significant difference ( $p=$ 0.284). This was further confirmed using the Bayes factor analysis done on both the accuracy and reading speeds using the different discs. It is interesting to note that, as seen in Table 3, the 12 BNB users needed only a single session of approximately 30 minutes each to reach more than 90% accuracy in character recognition unlike the BFS group who required 1–2 sessions on an average. It is reasonable to infer that it is due to their increased tactile experience. The session speed of the BNB users (6–7) wpm was similar to that of the BFS group.

4.4 Results-extended study

4.4.1 Results on increasing speed

Figure 10.

Highest speeds obtained during each test session TS.

Figure 11.

Session accuracy across each segment S1–S5 for each of the training sessions TS1–10.

It is seen from the results as shown in Fig. 9 that there seemed to be a plateauing effect during testing session TS8, TS9 and TS10, with the value at about 16 wpm. This is approximately 3.2 times his speed during the cross-sectional study using the same disc. It is higher than the mean reading speed of slow BB reader when using a single cell reader [26].

The highest number of words that were read during each test session were recorded. The highest speed among the 10 sessions was 19 wpm as seen in Fig. 10. The words presented to the user varied in length from 2 to 6 lettered words. These results suggest an increase in speed by the user over time.

4.4.2 Results on progressive desensitization

The method of reading introduced here involves the brushing of non-sensical characters under the users reading finger. Since this is a feature that is not seen in conventional braille reading, we analysed the effect of desensitization within each of the test sessions TS1–TS10 which would be indicated by diminishing accuracy across the segments S1–S5 of each of the ten sessions. The results as seen in Fig. 11 do not show a decrease in accuracy at from segment 4 to segment S5 except in the training session TS4 where the accuracy dropped from 100% to 90%. The user, JT, also verbally reported no sense of desensitization in his finger even though it was held over the rotating disc for the entire length of 20 minutes. He indicated that words of longer length would require a higher memory load.

5. Discussion

With the 17 BB users, a statistical analysis on accuracy between session with the prototype and paper braille showed no significant difference ( $p>$ 0.05) in accuracy within the same time frame of 1 minute. Further corroboration with the Bayes factor analysis supported the same. This fact strongly suggests the feasibility of using this method of reading. The accuracy measures of the forty users were high at 98.6% for blind braille readers, 90% for blindfolded sighted users and 96.8% for blind non-braille users. There seem to be no Stroop costs [27] among the different sets of users in terms of accuracy. Stroop effects occur when the reading of the word and the reading of print colour of the target are inconsistent.

On testing with non-braille users, the results showed no significant difference in accuracy of identification of words with both types of discs. These results concur with that of Heller’s [22] where it was found that between the sizes of 8 mm–11 mm, characters formed out of dots or lines were perceived similarly. In this paper, at sizes of 9 mm, there seem to be no significant difference in the accuracy of reading words with lined or dotted discs. Late-blind users undoubtedly have a richer tactile experience than the blindfolded sighted users which is reflected in the higher accuracy measures. Yet blind-folded subjects also have an active memory of the shapes of letters as they actively engage with it throughout the day.

It is seen that 40 characters would be needed for the simplest reading (A–Z, 0–9, ,.,?). A braille disc incorporating these 40 characters would need a diameter of 8 cm. The lettered disc would be of bigger diameter as the embossed letters were bigger than their braille equivalents. A slight decrease in the dimensions of the English letters can be made, but it would require further research to see the effectiveness using this similar method of assimilating words. In [22] it is seen that the letter sizes needed to be between 8 mm and 11 mm for tactual perception. The embossed English letters used in the present study were of size 9 mm and can therefore be further reduced. In embossing 40 characters along the circumference, a greater number of nonsensical characters would pass under the reading finger but since all the forty users reported no hindrance with this feature, we do not envisage a problem there. This method of presentation holds promise in terms of the design of a dual tactile reader which can be used by both the early-blind and the late-blind. It is appropriate to keep in mind certain caveats here-the embossed braille disc displays Grade 1 Braille and the lettered disc embosses Uppercase letters only as that is more perceptible to the user based on prior literature [21].

The speed reported in the experiments can be considered to be the minimum speeds at which they would read considering that they are not in full control of the session. The session speed reported includes the time delay between the experimenter responding to the answers of the user. Yet with the BFS users, the session speed was 6–7 wpm after just 1–2 training sessions. On comparing these results to [24] where it reported that blindfolded sighted users after 9 months of training could read braille at the rate of 6–7 wpm, it can be inferred that the memory of the shapes of the letters have contributed to a substantial reduction in training time needed to read tactually. Further testing with BNB users showed that just a single session was needed to get similar results. This reduction in training time strongly suggests that tactile readers must include raised English letters for the benefit of those who are late-blind and do not know braille.

The extended study done on a single BFS user showed an increase in speed over time. The final speed obtained after 10 test sessions was 16.4 which was more than the average speed of 12 wpm for slow BB readers when using single cell readers [26]. The extended study indicates that progressive desensitization caused by the rotating disc did not take place since the accuracy in reading words did not diminish. Even though the extended study was done on a BFS user, the results are reflective of similar trends for BB users i.e. higher reading speeds can be expected over time due to their increased tactile acuity. Whether there would be a marked increase in speed for a BB user would require further research.

The users were asked to rate the method of reading as difficult or easy. All of them rated it as easy to use. Two experienced users of screen readers said it was slower. Screen readers read out words at 300 wpm. The users who are now accustomed to such speeds felt even a full line of refreshable braille text would not match the screen reader in terms of speed. Possibly only a full page of refreshable braille would compete with the speed of screen reader software. Yet these are two different modalities of receiving information and should be viewed as complementary technologies instead of one replacing the other. How each user would prefer to receive information may vary. In the experiments here, some of the volunteers who used screen readers mentioned that they preferred tactile reading to hearing as it helped them to focus. A Blind Braille instructor who was a skilled user of screen reader software was of the opinion that it would definitely be beneficial to young students learning Braille. The non-Braille Blind university student stated that the English letter embossed disc would be very useful to non-braille blind users as a tactual mode of receiving information.

6. Conclusion

The design of a tactile reader where words could be presented using a single actuator has been explored. Since there is a substantial decrease in the number of actuators needed to display text, presenting words with a single actuator reduces costs. The design presented here allows for sliding indentation between the text and the finger pad. It was seen that words could be read even as non sensical letters brushed under the reading finger and letters were presented at varying refresh rates. The results obtained indicate that this technique of reading is viable. The benefit of this design is that it can cater to both early-blind and late-blind users. It is interesting to note that just a single training session was needed for each of the 12 BNB users. This short training time to read tactually is advantageous given the difficulty in learning Braille for late-blind users. While the speed of reading seemed comparatively slow in the cross-sectional study, an extended study on a single user suggests an increase in speed over time with reading speeds comparable to that of other single cell reading designs. While most late-blind users can use screen readers, designs that incorporate this reading method can enable them to read tactually from a PC. Which modality of receiving information they prefer would vary from person to person. Optimization of the size of the discs needs to be done so that it can hold all the characters needed for reading at reduced dimensions. This prototype had only 12 characters embossed on the braille and the lettered discs. In future, discs which holds all the letters of the alphabet for a more extensive study will be designed and developed. Nevertheless, the results of the present study provide information to guide the design of a single cell tactile display operating by lateral skin deformation, adaptable for use by both Braille and non-braille users.

Footnotes

Acknowledgments

This work was supported by The Department of Science and Technology, Government of India, under the Women’s Scientist Scheme (SR/WOS-A/ET-2014(G)). We are thankful to our volunteers, who so willingly gave their time for testing of the device. Thanks are due to Mr. Ravi Siva lingam for his assistance in the CAD drawings of the discs.

Conflict of interest

The authors declare that there is no competing interest.

Appendix

This gives the list of words used for testing with the embossed Braille Disc.

do	log	look	clock	no	dog
cook	gloom	good	cold	go	moon
god	lock	gloom	dog	con	look
rock	crook	gold	cook	hold	good

This gives the list of words used for testing with the embossed English letter disc.

IT	HEN	NO	HIDE	SHAKE	AS	DO	TOOK	SHOOK	IN
ANT	SACK	TIN	HOOK	CHIT	HITS	HAS	THIS	HAND	TOOK
KIND	DO	STITCH	THAT	THOSE	KID	COOK	KIND	AS	COT
CHAT	THESE	KIND	SET	KIDS	HOSE	SACK	NOD	CHIT	CHEAT
SAND	SKID	NOSE	NO	DIE	SHEET	SHOE	SHINE	SKIT	DOSE

This gives the list of words used for the Single User Extended study with the embossed English letter disc.

I CHOOSE TO DO THAT

HIDE THE SHOE IN THE DESK

HIDE THE HOSE

DANCE ON THE DESK

HIT AN ANT

I CHOSE TO EAT CHEESE AND CAKE

THE DAD CHASED THE CAT IN CHINA

KIDS EAT CHEESE

HIDE THE CHEESE IN THE DESK

10.

THE SHEET HAD A STAIN

11.

SHE CHOSE TO HIDE THE STAIN IN THE SHEET

12.

SHE IS CHOSEN TO TEACH

13.

SHE CANNOT STITCH THE SHEET

14.

STITCH THE HAT

15.

HOOK THE SHEET

16.

SHE CHOSE TO STITCH THE SHEET

17.

THE KITE NEEDS A HOOK

18.

SKATE IN THE HEAT

19.

SKID ON THE SHEET

20.

HIS HAND IS HOT

21.

HE SITS ON A HOT DESK

22.

HE HAS TEN SHOES

23.

SHE HAS ONE SHOE

24.

HIT THE DESK

25.

DO NOT CHEAT THE DAD

26.

SHAKE THE HOT HAT

27.

HIT THE HEN

28.

CHOOSE TO SKATE ON THE SAND

29.

CHASE THE CATS TO CHINA SOON

30.

THE SON CHEATED IN THE TEST

31.

SHE DID NOT CHEAT IN THE TESTS

32.

CAN HE SHOOT THE SNAKE

33.

I CANNOT HIDE THE SHOE IN THE HOT DESK

34.

THE SHOE IS HIDDEN IN THE SAND

35.

HE HAS CHOSEN TO DANCE IN THE SKIT

36.

DOES IT NEED A SHEET

37.

THE NEST NEEDS A STITCH

38.

I CAN SHOOT A SNAKE

39.

DO I CHOOSE TO SHAKE THE HAND

40.

IT IS HOT

41.

I CAN DANCE

42.

SHE CHOSE TO DO THAT

43.

I CHOOSE TO DO THAT

44.

HIDE THE SHOE IN THE DESK

45.

HIDE THE HOSE

46.

DANCE ON THE DESK

47.

HIT AN ANT

References

Leonardis

Loconsole

Frisoli

. A survey on innovative refreshable braille display technologies. In: Advances in Intelligent Systems and Computing. 2018; pp. 488–98. doi: 10.1007/978-3-319-60597-5_46.

Russomanno

O’Modhrain

Gillespie

Rodger

MWM

. Refreshing refreshable braille displays. IEEE Trans Haptics. 2015; 8(3): 287–9. doi: 10.1109/TOH.2015.2423492.

Vidal-Verdú

Hafez

. Graphical tactile displays for visually-impaired people. IEEE Trans Neural Syst Rehabil Eng. 2007; 15(1): 119–30. doi: 10.1109/TNSRE.2007.891375.

www.bristolbraille.co.uk [Internet]. Available from: http://www.bristolbraille.co.uk/.

Wright

Wormsley

Kamei-Hannan

. Hand movements and braille reading efficiency: data from the alphabetic braille and contracted braille study. J Vis Impair Blind. 2009; 103(10): 649–61. doi: 10.1177/0145482X0910301008.

Jones

. Designing effective refreshable braille displays. Computer (Long Beach Calif) [Internet]. 2016; 49(1): 14. doi: 10.1109/MC.2016.12.

Thomas

Rufus

. Alternate braille display designs: a review. Technol Disabil. 2017; 28(4): 123–32. doi: 10.3233/TAD-160451.

Runyan

Blazie

. EAP actuators aid the quest for the “Holy Braille” of tactile displays. Electroact Polym Actuators Devices 2010 [Internet]. 2010; 7642: 1–12. doi: 10.1117/12.847764.

Hribar

Deal

Pawluk

DTV

. Displaying braille and graphics with a “tactile mouse”. ASSETS’12, USA. 2012; (October 2012): 251–2. doi: 10.1145/2384916.2384978.

10.

Pawluk

Adams

Kitada

Member

Kitada

. Designing haptic assistive technology for individuals who are blind or visually impaired. IEEE Trans Haptics. 2015; 8(3): 258–78. doi: 10.1109/TOH.2015.2471300.

11.

Marc Hagen. The First Braille eReader

|

Closing The Gap [Internet]. 2018 [cited 2019 Jul 12]. Available from: https://www.closingthegap.com/the-first-braille-ereader/.

12.

Grunwald. Reading and writing machine using raised patterns. United States Patent, Pat No3624772. 1971.

13.

John W. Roberts, Damascus, MD (US); Oliver T. Slattery, Gaithersburg, MD (US); David W. Kardos C. REFRESHABLE BRAILLE READER. United States Patent, Patent No 6,776,619B1. 2004; 2(12).

14.

Slattery

Kardos

, Mulkens, Edwin SB. Apparatus and method utilizing b directional relative movement for refreshable tactile display. Pat No US 6,692,255 B2. 2004; 2(12).

15.

Wang

Hayward

. Compact, Portable, Modular, High-performance, DistributedTactile Transducer Device Based on Lateral Skin Deformation. In 2006 14th Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems. 2005 Mar 25; pp. 67–72. IEEE. doi: 10.1109/HAPTIC.2006.1627091.

16.

Lévesque

Pasquero

Hayward

. Braille display by lateral skin deformation with the STReSS2 tactile transducer. Proc – Second Jt EuroHaptics Conf Symp Haptic Interfaces Virtual Environ Teleoperator Syst World Haptics. 2007; pp. 115–20. doi: 10.1109/WHC.2007.25.

17.

Mahmoud Al-Qudsi. REFRESHABLE BRAILLE DISPLAY. Pub No US 2013/0203022 A1. 2013; 1(19).

18.

Tobin

Hill

. Is literacy for blind people under threat? Does braille have a future? Br J Vis Impair. 2015; 33(3): 239–50. doi: 10.1177/0264619615591866.

19.

Jernigan Institute. The Braille Literacy Crisis in America. Natl Fed Blind [Internet]. 2009; Available from: https://nfb.org/images/nfb/documents/pdf/braille_literacy_report_web.pdf.

20.

Merry

. Fingers for Eyes. The Story of Raised Print. Sci Mon, Publ by Am Assoc Adv Sci. 1937; 44(3): 273–9.

21.

Loomis

. On the tangibility of letters and braille. Percept Psychophys. 1981; 29(1): 37–46. doi: 10.3758/BF03198838.

22.

Heller

Nesbitt

Scrofano

Daniel

. Tactual recognition of embossed Morse code, letters, and braille. Bull Psychon Soc. 1990; 28(1): 11–3. doi: 10.3758/BF03337634.

23.

Thomas

Rufus

. Investigations on Laterotactile Braille Reading. In: Human-Computer Interaction – INTERACT 2017 Lecture Notes in Computer Science [Internet]. Mumbai, India: Springer; 2017; 196–204. doi: 10.1007/978-3-319-67744-6_13.

24.

Bola

Siuda-Krzywicka

Paplińska

Sumera

Hańczur

Szwed

. Braille in the sighted: teaching tactile reading to sighted adults. PLoS One. 2016; 11(5): 1–13. doi: 10.1371/journal.pone.0155394.

25.

Rouder

Speckman

Sun

Morey

Iverson

. Bayesian t tests for accepting and rejecting the null hypothesis. Psychon Bull Rev. 2009; 16(2): 225–37. doi: 10.3758/PBR.16.2.225.

26.

Ramstein

. Combining haptic and braille technologies: design issues and pilot study. In: Proc 2nd ACM Conf Assist Technol ASSETS 96. 1996; pp. 37–44. doi: 10.1145/228347.228355.

27.

Jarjoura

Karni

. A novel tactile braille-stroop test (TBSt). Br J Vis Impair. 2016; 34(1): 72–82. doi: 10.1177/0264619615612344.