Analysis and Evaluation of Braille to Text Conversion Methods

,


Introduction
Visually impaired people are an important part of every community [1]. ey are also concerned in learning the details of everything they encounter in their daily life [2]. e total number of visually impaired people is 2.2 billion; among them, 36 million are completely blind, and rest of the 1 billion have moderate to severe vision impairment [3].
Approximately thirty-seven million of the six billion populations worldwide are suffering from blindness. Unfortunately, 80% of blind people live in developing countries with restricted facilities for them [4].
Smartphones have become an integral part of everyday life. An expected increase of smartphone users will increase up to nine billion by 2022 [5]. e widespread use of smartphones has brought significant changes in how people learn. Research indicates that about one-third of smartphone usage consists of educational activities. Although smartphone usage has increased exponentially, it has low prevalence among people with visual disabilities. ere are many complex accessibility issues that must be resolved in order to enable the full inclusion of this community [6]. Accessibility issues have been an important research domain over the last few years promoting the development of thousands of smartphone applications to help people with a visual disability, e.g., voiceOver services, talkback services, screen readers, and navigators. ese researches resulted in a dramatic increase in mobile-screen reader usage for the visually impaired, from 12% in 2009 to 88% in 2019 [7]. Despite the benefits that smart devices can offer, if the learning applications are not properly designed, their touch-screen interfaces may place an extra burden on blind learners. ere are features such as VoiceOver for iPhone that help blind users interact with their device and browse content. However, educational applications often fail to consider the interaction patterns of blind learners with smart devices. e language visually impaired people use for reading and writing is known as Braille, which was designed by Louis Braille. It is composed of six raised dots that can be easily written by visually impaired [8]. His design is illustrated in Figure 1. Each Braille character is represented using a combination of six dots arranged in a 3 by 2 matrix [9]. e Braille code system has been widely adopted in several communities because of its simplicity and comfort. Braille has been supported by different languages such as English, Arabic, and Hindi, among others [10]. However, few studies have been conducted on Braille for smartphones.
Research on Braille to text conversion has been carried out in the USA, Canada, India, Pakistan, and France. e literature shows that majority of the conducted research is limited to the USA and Canada. is indicates that there is a considerable demand for such a study in the rest of the world [6].
With the advancement of technology, Braille scripting mechanisms became an important research domain. Within this category, an initial device called Perkins Brailler was introduced to facilitate Braille writing. Space, backspace, and line space keys were designed in Perkins Brailler, as well as keys corresponding with each of the six dots in the Braille code [11]. In 2008, a lighter and quieter version was developed and launched that included an erase key and integrated carrying handle, which was not available in Perkins Brailler.
Another adaptation of the Perkins Brailler, the SMART Brailler, was created by David S. Morgan and released in 2011 [12]. Along with the existing features, Smart Brailler also included text-to-speech functionality in several languages. With the advent of computers, many users created Braille output by connecting a computer and Braille embosser. Visually impaired users were able to read the computer screen using screen reader computer software and/or Braille displays. Another similar Braille recognition system was designed by [13]. In this scheme, images were distributed into three threshold values, and Braille characters were subsequently recognized. Effectively, this interpretation was used to create a suitable dictionary. Recent research has focused on eliminating the need for separate hardware in Braille scripting. Application-level software has been designed to facilitate Braille users.
is survey focuses on gathering the difficulties faced by visually impaired while using a computing-based Braille input mechanism. Many technology-oriented applications for the visually impaired are available. Only those applications are considered for these surveys that are part of research taking place in different countries. Studied applications were analyzed and compared based on matrices related to usability issues for touch-screen-based Braille input methods. ese evaluations bring forward the strengths and weaknesses of current schemes with a special focus on usability. No such study exists in the literature, so this paper can provide guidelines to the researchers for designing future applications that are highly usable for visually impaired people. is paper comprises the following sections. Section 2 gives a detail insight into the previous studies that have been designed for entering Braille data. Section 3 describes the methodology in detail. Input methods are compared and evaluated in Section 4. Usability issues gathered from different studies are also included in this section. Section 5 concludes the paper and gives future recommendations.

Previous Work
is survey paper provides a review of the current state-ofthe-art Braille input methods. In this section, we provide a detailed insight into the problems with these schemes when these are used by visually impaired people. Based on the current survey, we have identified new directions for future research. In recent years, many studies have been conducted to make Braille more technology-assisted. To analyze these studies, we have broadly divided the Braille input mechanism into two main categories: In the scanned Braille input, Braille Dots are extracted from Braille sheets using a scanner and, then, converted into text using optical character recognition, as shown in Figure 2. In this mechanism, visually impaired users give input on sheets without any interaction with a computing device.

Arabic Optical Braille Recognition System.
A study was conducted that takes input from a flatbed scanner, as it clearly displayed the Braille Dots. e scanned image was converted into grayscale, the image frame was cropped, and the resulting image was stored in a 2D array. To remove the skewness in the framed image, an algorithm was designed. Finally, the Braille cells were recognized. ey achieved approximately 99% accuracy for Braille written in Arabic with single-and double-sided scanned documents. ey did not evaluate their system against any other application [14].

Deterministic Turing Machine for Context-Sensitive
Translation. To reduce the written communication gap between visually impaired and sighted people, a mechanism was developed to convert Braille codes into Urdu text [15]. ey used a scanned Braille document as the input, which was, then, converted into a binary grayscale image. Multiple experiments were conducted for adjusting the threshold value of the scanned document. e primary focus of this study was to evaluate the context-sensitivity of Urdu- Braille. e main issue in the development of Urdu-Braille is that Braille is written left-to-right, whereas Urdu is written rightto-left. is issue was resolved by inventing a deterministic Turing machine that translates Urdu into Unicode and, then, reads the Unicode from left-to-right.

Text Translation of a Scanned Hindi Document into
Braille via Image Processing. Hindi to Braille conversion was performed so that visually impaired people could have access to a wider range of Hindi literature [16]. ey designed a database for consonant and matras generation using an image segmentation technique. Segmented letters were matched with the generated Hindi database using Principal Component Analysis before conversion into equivalent Braille codes. Braille cells for the Hindi text were generated with the help of mapping tables. Every image in the database was sized to 187 × 128 pixels to minimize the memory requirement. is scheme also requires less programming time as it utilizes a letter position in the Hindi database for letter matching.

Braille Instant Translator.
e Braille Instant Translator converts Braille documents into the English language by identifying Braille Dots embossed on a page [17]. e scanned Braille document uses a camera-generated image as input and identifies the Braille Dots to emboss on the page in a grayscale image format. e extracted image was enhanced through standard deviation evaluation. A histogram was generated from the pixel values for image binarization. e brighter regions were assigned a value of 1, and the remaining regions a value of 0. MATLAB image-processing techniques were used to identify dots that were, then, matched with the English alphabet. Effectually, results indicated that the study achieved 80% accuracy, but implications are limited to only Grade 1 Braille.

Web-Based Urdu-Braille Translator. A web-based
Urdu-Braille translator was developed by Iqbal et al. [18] for parents of the visually impaired. e translated Urdu script can be exported to PDF or directly sent to an embosser. Four modules were developed to enable an explicit learning process: a multiplechoice interface, an Urdu-Braille Reading module, a Braille-to-Urdu word mapping module, and a fillin-the-blank module for Braille-to-Urdu word translation. To prove the effectiveness of their Urdu-Braille translator, a usability study was conducted on the parents of 15 blind people. For these tests, the participants used the Urdu script from BBC Urdu and translated it with the Urdu-Braille translator. Results indicated that, with each new test, the number of correct answers increased, verifying the effectiveness of the translator when applied to Braille learning. Given the educational impacts of the translated Urdu script, parents were able to use the online translator's Braillelearning application for more involvement with their visually impaired kids.
2.1.6. Feature Extraction Using SDAE. Braille pattern feature extraction is a cumbersome task. To reduce complexity, Stacked Denoising Auto Encoder (SDAE) was applied for Braille recognition [19]. SDAE is a deep learning technique used for automatic feature extraction and dimension reduction. To create the initial dataset, the authors produced a sample dataset of Braille images by segmenting the original photos taken with a digital camera from a Braille book. is scheme used SDAE for Braille feature extraction in order to obtain initial weights. e output of SDAE was used as input in SoftMax to build the classifier for training, which employed a traditional supervised learning algorithm with an initial dataset. Consequently, a deep network was initialized, and then, weights were fine-tuned using a backpropagation algorithm. A comparison of three different networks was performed, including multilayer perceptron (MLP), radial basis function (RBF), and SoftMax. Results indicate that, in Braille   Mobile Information Systems before being converted into the binary and speech format. e authors were able to achieve an efficiency of 99%. It took only 26 seconds to translate a two-sided document into text and speech.
2.1.8. ODIA Braille. Another application named ODIA (a language used in East India) Braille was designed by the authors [21]. In this scheme, a scanned Braille document was converted into text and vice versa using the image-processing technique. Initially, a greyscale image is acquired from the scanned document. is image is enhanced using feature extraction and Braille cell segmentation technique. After this, recognized patterns are stored in the database.
is database is verified by using different Braille samples for conversion.

A Braille Recognition System by Using the Mobile Phone with an Embedded Camera.
A mobile camera-based application is designed in [22]which captures a Braille image. Using this application, a visually impaired person can convert a Braille written text anywhere easily. Noise removal is performed to extract the important features from the image using segmentation and fast dynamic thresholding techniques. A grid, constituting the location of the dots, is converted into vector form, from which Braille characters are recognized and translated into English characters. In this scheme, the time required to convert an image depends upon the quality of the image. Images with more noise will take a longer time to process. is application guarantees a 100% noise reduction. After noise removal, it only takes 2 seconds for the conversion process.

Recognition of Ethiopic Braille Characters.
A Mechanism to recognize Ethiopic Braille characters was designed in [23]. e authors designed a new skewness correction technique. In the first step, noise is removed, and then, segmentation is performed using the direction field to detect the exact regions of the Braille Dots. In the next step, skewness correction is performed. After this, important Braille cell values such as the height of the cell, the width of the cell distance between different cells, and Braille character lines are identified. Initially, half characters are detected and recognized. en complete Braille cell formulation and translation is performed. ey tested their prototype in MATLAB. An average accuracy of 96.5%-98.5% is achieved for poor to medium quality images, while 99.9% accuracy is achieved for good quality images. Currently, this scheme can only convert one-sided Braille documents. Table 1 provides a comparison of these schemes for which the Braille input is extracted from scanned documents or camera images.

Touch Screen.
In this method, Braille data is input using a touch screen. As outlined in Figure 3, these images are compared with the Braille dataset. e matched Braille character is then conveyed to the user using various output techniques.

NavTap and Braille Tap.
e NavTouch keyboard for mobile devices was developed by Guerreiro et al. [24]. is layout of the keyboard divides the alphabet into five rows, starting with a vowel each row. By performing navigation gestures in four different directions, users can navigate through these rows: up, down, left, and right. Both vertical and horizontal navigations are cyclical. Audio feedback is also provided in order to locate the desired letter. ree groups of five users, who had no previous experience in mobile text entry, evaluated designs. Users were first trained on the text entry method. Results indicated that NavTap outperformed other layouts. However, these keyboards require both hands for operation, which is difficult for visually impaired people.

No-Look Notes. Bonner designed No-Look
Notes an eyes-free, gesture-based text entry system for multitouch devices. In this layout, characters are arranged on the screen in eight pie-shaped menus, in which each segment includes three to four alphabets that correspond to the international standard mapping a phone keypad to letters. A user can move his or her finger on the pie menu, and the voice feedback will respond to assist them in selecting the appropriate character. Gestures are used to enter spaces or undo any action. is layout was tested on a group of 10 visually impaired users and evaluated based on matrices, such as words-per-minute (wpm), as well as the relationship between speed, accuracy, and errors. Results indicated a 100% increase in wpm for No-Look Note as compared with VoiceOver, but both exhibited approximately the same error rate. Not only is this scheme specific to English text entry but also to users complaining about fatigue during the entry procedure [25].

V-Braille.
V-Braille is a method that conveys Braille characters in mobile devices using vibrations [26]. In V-Braille, the screen is divided into six portions of 3 rows and 2 columns. Each area represents a single Braille Dot. e phone vibrates when the user touches a region that corresponds to the raised dots in the current character. For example, touching the area for dot 2 and 5 provides a strong vibration to help users identify screen dots that are vertically adjacent. V-Braille is a useful output method for deaf-blind mobile device users, since V-Braille only provides haptic feedback. To evaluate this design, a user study was conducted. ere were 6 male and 3 female participants in the study. e first task assigned to the users was reading 10 random characters, and the second task was reading short sentences. e average time calculated for reading a single character was between 4.2 and 26.6 seconds. More than 50% of the users were able to read the characters in less than 10 seconds. e time calculated for reading a short sentence was 130-781 seconds. More than 70% of participants clearly understood the meaning of the sentence. Subsequently, a semistructured interview was also conducted to elaborate on the overall environment of the application. Participants reported feeling happy and relaxed using this application. Designers of Braille Play games [27] built upon V-Braille, adding a speech output that tells the users which dot they are touching. ey also created a V-Braille interface for entering characters, where the user double taps regions of the screen to raise the corresponding dots. e TypeIn Braille keyboard uses gestures to enter Braille characters. ree different gestures of singleand multitouch were used for activating and deactivating Braille Dots. Audio or vibration feedback confirms the user input. Flick gestures were defined for input and editing options, such as the end of character, space insertion, text deletion, capitalization, text selection, and cursor relocation. Gesture memorization, time consumption for moving between writing, exploration, and selection mode are the basic drawbacks of the system [28].

Braille Touch.
A study conducted on Braille keyboard design used six specific screen locations on touch-screen devices for entry of the Braille code [29]. Data entry required both of the user's hands, specifically three fingers from each hand for entry of a single character. Users faced problems simultaneously operating multiple fingers and both hands. To operate this keyboard, blind users must distance the phone from their faces, which can be uncomfortable and lead to a breach of privacy. Users were forced to write on the fixed locations mentioned on the screen, which burdens visually impaired people who struggle to keep track of the location. Several usability experts evaluated Braille Touch and two visually impaired users. e most common feedback was to provide a position-free text entry that did not require distance between users and their phones. is keyboard was designed to avoid the multitouch technique [30]. Users must perform a single touch to enter a Braille Dot, and after a period, the user receives an audio confirmation. e timer for audio feedback can be adjusted per user's expertise. e user must double-tap the screen anywhere to accept the entered Braille character. e left swipe gesture is used to clear marked Braille cells. ey evaluated Braille Type on 15 blind users. Results indicated a text entry speed of 1.45 wpm, a significant improvement from the 2.11 wpm achieved by VoiceOver. e error rate of Braille Type was comparatively low at 8.91% as opposed to VoiceOver's 14.12%. Overall results indicated that Braille Type, when compared with VoiceOver, was easy to learn and exhibited a low error rate, but performed slower. Due to less screen space, the integration of this keyboard with other applications is not possible.

Mobile Brailler.
A new keyboard layout was designed based on the similar design patterns to those in TypeIn Braille [31]. Along with the tapping function, this keyboard also used swiping gestures for inserting and deleting characters. To enter two dots in a consecutive row, the user must tap with two fingers at the same time. To deactivate the left and right dots, a user must swipe to the right, which removes the last entered character. Mobile Brailler prototyped and compared five different input methods: One-Finger, Split-Tap, Two-Finger, umb-Typing, and Nine-Digit. Along  Mobile Information Systems with the tapping function, the right and left swipes are used for adding spaces and backspaces, respectively. e study was conducted on 15 visually impaired people from the greater New York City area, and the prototypes were tested on the android environment. Users answered qualitative and open-ended questions in each application. Results indicated that the One-Finger method was preferred by the visually impaired because of its simplicity. Gesture memorization was among the primary drawbacks. e typing speed of other techniques was faster than the single-finger method, but those techniques were not preferred due to gesture difficulty. Recommendations for these input methods were provided based on the user experience in order to improve prototype deficiencies.

Braille Key.
Braille Key was designed with four large keys in the screen corners. Two buttons on the upper side are used to enter two columns of Braille code. A user must perform a single touch for entering the first dot, a double-tap to enter the second dot, and a long tap to enter the third dot. e lower two buttons are used for editing the text. e developed prototype was compared with Apple's wellestablished VoiceOver. Five visually impaired people participated in the study. Users were given a brief 10-15 minute awareness session on how to use both of the applications. Two sentences were entered without correcting any mistakes. Text entry speed and typing accuracy were measured for both applications. Effectively, Braille Key text entry speed and accuracy outperformed the iPhone's VoiceOver. e primary limitation of Braille Key is the identification of the button position on the screen [32].

Perkinput. Perkinput is a novel technique designed by
Azenkot that uses the input finger detection (IFD) method. e input is entered into the device through multipoint touch. Tracking algorithms are used to detect the input finger reference point [33]. is method uses a 6-bit Braille code with audio feedback and provides single-and doublehand options. A study was conducted on eight users to evaluate this input method. Results found that Perkinput outperformed the iPhone's VoiceOver in speed and accuracy. A case study was conducted for performance evaluation, confirming that Perkinput improved writing proficiency without errors. e Perkinput keyboard is particularly advantageous because it eliminates the fixed-key concept, successfully resolving navigational problems. However, creating a reference point requires users to single, double, and triple tap, the latter of which is time-consuming and often results in users forgetting the location of the reference point.

Braille Calculator.
Learning mathematics is as important for the blind as it is for the sighted people. e Braille Calculator is designed so that visually impaired people can easily learn mathematics [34]. e input is taken from a touch-screen device. e Braille Calculator uses a 4-wire resistive 2.8-inch touch-screen. e screen is divided into six portions for text entry. e touch-screen is interfaced with the Atmega 328 microcontrollers. An analog to digital converter (ADC) gives the user's impression of the location coordinates. e user is provided with an audio feedback in response to the dot entered. A step-by-step input is received for solving complex equations. Finally, the user is provided audio feedback regarding the solution. e Braille Calculator uses the Atmega 328 microcontrollers and a Secure Digital Card (SD) interface.

2.2.11
. VB Ghost. VB Ghost, based on the Ghost word game, is an educational smartphone game for people with low or no vision [35]. e V-Braille interface was used as the basis of this game. e game was specially designed for educational and recreational purposes. is application was developed to reduce the accessibility issues that arise while using Braille on touch screens.
Both the hepatic and audio feedback were provided. Vibration occurs at the place where a raised dot occurs. In this game, a word fragment is presented and the player has to complete it. When satisfied with their letters, a player can submit the word by pressing the enter key or swiping with two fingers. e primary purpose of this game is to demonstrate the potential of developing fun, accessible, and educational games for visually impaired users.

Braille Play. VB Ghost was further improved and developed in the form of a complete suite named Braille
Play. e suite consists of four different games: VBReader, VBWriter, VBHangman, and VBGhost. A longitudinal study conducted resulted in only one child capable of playing the game independently. However, some children were able to acquire the basic Braille-learning skills [27].
To analyze accessibility issues for blind people, four different smartphone applications were evaluated, namely, Blind Navigator, Easy Phone for the Blind, Blind Launcher, and Call Dialer [36]. To perform the study, ten visually impaired people were selected from various educational institutions. Ease of use, learning ability, absence of errors, efficiency, and voice understanding were the primary matrices for comparison. Survey forms were used for collecting feedback. Results indicated that, currently, most visually impaired people use Symbian phones, but blind people who already use smartphones are not ready to use any other device. Furthermore, 70% of the blind people easily understood the message in their Pakistani native language.

Eye Droid Keyboard.
Another keyboard, Eye droid, was designed for entering Braille patterns using different gestures. To calculate the minimum swipe distance, input coordinates of the Braille Dots entered, that is, X1, Y1 and X2, Y2, are extracted, and the swipe threshold velocity is calculated using Velocity-X and Velocity-Y. e different gestures employed were left-to-right, right-to-left, and bottom-to-top swiping, as well as screen tapping [37]. ese gestures, correspond with functions including activation and deactivation of left and right dots, activation of both dots, and activation of a single dot. e subsequent Eye droid-B scheme was compared with the earlier Eye droid-A design. Survey participants found Eye droid-B to be faster and easier to use when compared with Eye droid-A.
is keyboard resolved navigational problems by eliminating the locationspecific buttons that troubled users. Alternatively, users only need to memorize the defined gestures.
2.2.14. Edge Braille. Edge Braille introduces another text entry method that designed buttons in the touch-screen corners. is method allows a user to draw a continuous line by swiping along the screen edges for the entry of a specific character. When the user slides his or her finger on the screen, a vibrio tactile and voice feedback is returned to inform the user of Braille Dot activation or deactivation [38]. e input speed of Edge Braille was compared with TypeIn Braille and Perkinput. Results showed that the input speed of Edge Braille was faster than that of Braille Type and slower than that of Braille Touch. Furthermore, users found the Edge Braille interface easy to operate. However, numerous problems still exist with this design, since it is impossible to draw lines for all characters based on dot position. is method limits users to enter only alphabets and numbers and does not allow for "editing." Additionally, activation and deactivation of different dots reduces application speed.

Braille Easy.
Braille Easy was developed for the entry of Arabic and English Braille codes within a mobile application. is system also used gestures, but for activation of the first and second column, users are required to tap once, twice, and three times [39]. is keyboard was significantly difficult for users to operate because it requires the memorization of different reference points. e keyboard speed was evaluated at 7 wpm, but the error rate was not specified. Furthermore, the keyboard only supports Grade 1 Braille.

Braille Ecran.
A tactile interface cover for touchscreen phones was designed by [40]. is cover, called Braille Ecran, which provides an interface that consists of six Braille Dots. Each dot was given a tangible button that users can find and press. e significant advantages of this design include the "editing" capability, as well as vibration and audio feedback. However, users encountered considerable problems with the system, including confusion when operating buttons due to a single key's multifunctionality. Additionally, data entry was associated with a high probability of error due to the close proximity of buttons.

Single-Tap Braille.
Single-Tap Braille is a positionfree text entry method. A user can enter text, numbers, and punctuation by tapping anywhere on the screen. An algorithm runs in the background, interpreting the user's finger tapping for the identification of the specific corresponding character. e significant limitation of this model is the location-specific memory required; blinds often struggle to remember where they tapped once they have picked up their finger, eliminating their ability to complete subsequent actions correctly. No audio or vibrotactile feedback was provided in this scheme [41].

Braille Enter.
Considering the problems highlighted in Single-Tap Braille, Braille Enter was designed to improve upon the method, consequently reducing on-screen navigation problems. In this method, the text is entered by activating and deactivating the Braille six-dot pattern, as shown in Figure 4. Additionally, the model allows for singlehand use and supports the entry of upper and lower-case letters, numbers, and special characters. Special functions are also available, such as adding and removing spaces. Furthermore, audio feedback was also provided to the users. e input received was used for activation and deactivation of the Braille Dots [42]. e press gesture entered the active dots, and a long tap deactivated the dots. e swipe function changed the character mode. e primary problem with Braille Enter is that users must enter all six dots even if only one dot is necessary, resulting in an excessively time-consuming process.

Braille Sketch.
Braille Sketch, a gesture-based input method, was designed for visually impaired users on touchscreen devices [43]. A user simply draws a gesture for text entry. Audio feedback is provided when words are completed as opposed to letters to reduce time consumption. For error correction typing, an auto typing algorithm was used. A study was conducted on ten participants with visual impairments to evaluate the method. Each participant completed five typing sessions, and results demonstrated that Braille Sketch supports a text entry speed of 14.53 wpm with a 10.6% error rate.
In summary, in contrast to the immediate letter-level, Braille Sketch provides audio feedback to encourage users to type more quickly. To correct typing errors, an autocorrection algorithm is used.

Methodology
e research problem that motivated the conduction of this study was to highlight the usability issues faced by visually impaired people while using the latest technologies for Braille writing. is study also highlights a deeper understanding of which methods are in use for converting Braille into natural languages to enhance the scope of work performed in Braille language.
is is a survey-based research in which several searches were made to collect relevant research articles. ese articles were collected from authentic resources such as Web of Science, IEEE Xplore, and Springer. Different queries were placed for searching such as the "Braille input method," "touchscreen-based Braille input method," and "touch-screen-based input method for visually impaired people." After skimming the results, only those papers were selected that take input from the visually impaired in Braille language, and then, they were processed into natural language. ese research papers were broadly categorized as scanned-based and touch-screen-based on their input mechanism. In scanned-based, Braille sheets are scanned and input is given to the computer for further processing, while touch-screen-based directly took input from the visually impaired on touch-screen devices and, then, converted that input into its equivalent character.
After categorizing, a visit was made to the National Special Education Center (NSEC) "Mannak Payyan" which is the only local school for people with impairments. Both methods were discussed with the students of the National Special Education Center. For scanned input, since the Braille is written on paper, the students felt no difference in using computing technologies. Touch-screen-based input provided more motivation for the use of the latest technology. us, questions related to general user experience on entering Braille on touch screens were asked. A theoretical output was made after analyzing this raw data from which concern problems experienced by the visually impaired people were extracted and used as matrices for evaluating the usability issues. erefore, we can design a system that helps the visually impaired community by fulfilling all their needs.

Comparative Analysis
A performance analysis was performed against VoiceOver based on the input efficiency of different techniques. On average, users require sixteen seconds to enter one word (approximately five Braille characters) using VoiceOver, which was designated as the standard [30].
In sixteen seconds, each method achieved the following word entries: Braille Type with 0.687 words, TypeIn with 1.211 words, Perkinput with 1.516 words, Braille Key with 0.524 words, and Edge Braille with 1.14 words. Evidently, TypeIn Braille and Perkinput outperform the other schemes with respect to data entry speed, as illustrated in Figure 5. Table 2 summaries the efficiency and accuracy of Braille input schemes currently in use.
Our analysis found that Edge Braille exhibits not only the highest input efficiency at 7.17 wpm but also the lowest accuracy. Alternatively, V-Braille exhibits a low input efficiency (1.32 wpm) but achieves 90% accuracy. Braille Enter exhibits a similar pattern, achieving a text entry speed of 2.45 wpm with 85.88% accuracy.
ese results indicate that, in order to achieve high accuracy, input efficiency is compromised. Research is needed to`design a Braille input technique that delivers satisfactory efficiency along with high accuracy.
We also analyzed how age affects learning Braille on touch screens. Most of the devices used for the assessment ran an android operating system. Figure 6 demonstrates that the age of the participant does not affect the learning time.

Usability Analysis.
e usability analysis of touchscreen-based Braille input mechanisms is summarized in Table 3.
is analysis was performed on different input schemes based on factors such as tool learning complexity, ease of use, feedback, language support, screen location dependency, and gestures used. Furthermore, two types of feedback, audio, and tactile were explored.
English is the most commonly used language in the current schemes. Visually impaired users found that applications that did not force them to touch specific locations for Braille Dot entry were easy to use. Alternatively, schemes requiring multitouch or memorization of a large number of gestures were found difficult to use.
All of these schemes were designed to enable visually impaired members of the society active participation in the advancing technology. ese input methods focus on facilitating touch-screen Braille input for blind users. e primary disadvantages of the current schemes are discussed in the following sections.

Screen Location Identification.
Keeping track of a specific location on a touch-screen device for entry of Braille Dots is a tiresome task for visually impaired users. Among these schemes, Edge Braille was considered the easiest to use, as the visually impaired were better able to identify the edges of the device.

Screen Location Identification.
Keeping track of a specific location on a touch-screen device for entry of Braille Dots is a tiresome task for visually impaired users. Among these schemes, Edge Braille was considered the easiest to use, as the visually impaired were better able to identify the edges of the device.

Use of Both Hands.
Around 30% of the touch-screenbased schemes studied in this survey require that the visually impaired use both hands for entering Braille characters [24,29,31,39]. is is a cumbersome task, especially if a blind user is walking or multitasking.

Multitaps.
Long press, double-tap, and triple tap each require advanced usability of touch-screen devices. e following schemes require the use of multitaps for entering Braille text [28,32,38,39,42]. ese mechanisms were found confusing to blind users, as mentioned by the examined literature.

Use of Gestures.
e utilization of too many gestures significantly increases the Braille keyboard learning time, which was especially evident in [27,28,37,42]. Specifically, mapping tasks with gestures, along with entering Braille Dots, caused a significantly higher error rate when tested on visually impaired users. e provision of an eyes-free, comfortable text entry method is a vital research area. In current schemes, all the burden of correct Braille entry is placed on the visually impaired user. ere is a substantial need for a mechanism that moves the burden from the visually impaired users to the technology, enabling greater accessibility and usability for this specific community.
On the basis of usability analysis, we have designed four new categories that can help researchers to design a better application for the visually impaired.
(i) Interactive display: for a visually impaired user, an interactive display would be one that is not screenspecific and also provides feedback. It could be an auditory feedback or tactile feedback (ii) Efficiency of use: people with visual impairments find those applications more attractive that are less     [43] complex, can be handled using a single hand, and are simple to use. (iii) Memorability: location-free specific applications or application that needs few gestures to remember are more appreciable by the visually impaired people. (iv) Recovery from errors: applications that allow editing or reentering text helps in recovering from errors. Table 4 presents the current schemes as per new categorization.

Conclusions
is survey paper focused on technological assistance available for visually impaired people. Braille input mechanisms can be categorically divided as scanned-and touchscreen-based input methods. In the scanned input, handwritten Braille sheets are scanned using scanners. Various studies have applied machine learning techniques such as optical Braille recognition, deterministic Turing machine for context-sensitive translation, feature extraction, and image processing to extract Braille Dots and convert them into a specific language. In the touch-screen-based input method, Braille Dots are entered using a touch screen on handheld devices such as mobile phones or tablets. Braille consists of six dots, and the basic mechanism of entering Braille using a touch screen requires entering Braille Dots by activating and deactivating pixels on the screen. Once the input is acquired, various algorithms process the extracted Braille Dots and convert them into their equivalent natural language characters or words. ese touch-screen methods use haptic, audio, and tactile feedback to assist visually impaired users.
is study compared different input methods on the basis of the entry speed and accuracies achieved, techniques used in the input methods, the number of participants on which the study has been conducted, gestures used, usability level, language used, feedback provided, and screen location independency. ese comparisons enabled us to understand the strengths and weaknesses of the current applications. Based on this literature review, we plan to design an application that provides high usability to the visually impaired students. In the future, a newly designed application can be compared with the previous techniques to improve its performance. Machine learning techniques can be applied for acquiring better accuracy for Braille to text conversion.

Data Availability
Data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this paper.