Abstract— Braille document
recognition involves image acquisition stage, image preprocessing,
binarization, enhancement, feature extraction and recognition of the braille
script from the given image. The paper concentrates on various methods for the
conversion from Southern Indian Braille script to corresponding language. The
proposed methods were based on histogram technique, Piece wise enhancement
techniques, Image filtering, horizontal profiling, vertical profiling, image
Thresholding, de-skewing the image, template matching etc. This paper discusses
research efforts proposed earlier for recognition the Southern Indian braille
script from a Braille document.
Keywords—Braille recognition, Image processing, Southern Indian Braille System, Visually Challenged
on Braille document recognition plays one of the hot research areas. The Braille encoding system
represent textual documents in a readable format for the visually challenged
persons. As there is a shortage of Braille compatible reading materials,
visually challenged people face trouble in necessities like education and
employment. Also, there are many important Braille Script documents that are
required to be printed so that they can be preserved and retrieved. There is a
need of a system for automatic recognition of Braille documents to reduce communication gap between the written text
systems used by sighted persons and access mechanisms through which visually impaired people can communicate.
1,2 is a tactile writing system used by visually challenged people. Braille
is a system of raised dots that can be read with the fingers by people who have
low vision or blind. Braille is named after Louis Braille, the French man who designed
it. Braille symbols are formed within units of space known as braille cells. A complete
braille cell comprises of six raised dots arranged in two parallel rows each
having three dots. The locations of dot are recognized by numbers from one
through six and it is shown in the figure 1. Sixty-four combinations (2^6) are
possible using one or more of these six dots. A single cell can be used to denote
an alphabet letter, number, punctuation mark, or even an entire word.
1 braille translates braille by changing the letter with the braille character and
is generally used by the beginners. The disadvantage of Grade 1 braille is that
braille characters are larger than ordinary letter. Grade 2 braille are
contractions and it permits to save space and increase reading speed. But translating
a text into Grade 2 braille needs special training and education. Grade 3
Braille includes many more extra contractions. It is rarely used for books. Grade
4 braille script are used by very few people. It is a shorthand script. Grade 4
script are used by blind people as a shorthand to follow oral conversation.
for Braille Embossed on Paper
are standards for the size and spacing of braille script 3. The braille dot’s
height is around 0.48 mm and will be uniform within any specified copy. The base
diameter of braille dots would be around 1.44 mm. Braille dot cell’s spacing should
follow the following: i) the distance from center to center of adjacent dots
(horizontally or vertically, but not diagonally) in the same cell is 2.340 mm ii)
the distance from center to center of corresponding dots in adjacent cells is
6.2 mm. The braille cell’s line spacing from center to center of nearest
corresponding dots in adjacent lines shall be 1.0 cm. Braille cell dimensions
(in inches) is shown in the figure 2.
Indian Language Braille System
languages are a language spoken mainly in Tamil
Nadu, Kerala, Telangana, Andhra, Karnataka and as well as in Bangladesh , Bhutan ,
Mauritius, Sri Lanka, some parts of Pakistan, Burma, southern Afghanistan, Nepal,
Africa, Indonesia and Singapore.
The Dravidian languages with the most speakers are Tamil, Kannada ,
is used by many people all over the world in their mother tongue, and gives a
means of knowledge for all. Bharati braille 4 or
Indian braille is a largely unified braille script
for writing the Indian
languages. Initially 11 braille scripts were in usage in different area
of the India and for different languages. Bharati braille had become a
nationwide standard braille script and it has been followed by Bangladesh., Sri
Lanka, Nepal. Figure 3, Figure 4,
Figure 5 and Figure 6 shows the Tamil, Telugu, Malayalam and Kannada braille
alphabets sheets respectively.
Southern Indian Language Braille
System Recognition Techniques
recognition involves image capturing stage, preprocessing,
binarization, enhancement, feature extraction and recognition of the
braille script from the given braille image. Earlier works on Braille character
recognition shows that there are very few research works on recognition of
southern Indian language Braille characters. Recognition of English Braille
script to corresponding English apathetes is easy compared to Southern Indian
languages as there are more than 260 characters in Dravidian languages.
Santhoosh et.al 5 presented
a research to Tamil braille character recognition based on camera assistive device, an embedded system bulit
on Raspberry Pi board. As a first step the captured image was converted to gray
image and the image was cropped according to the requirement. Adaptive
thresholding technique was used to separate the Braille dots from the
background. Morphology techniques were used to enhance the image and binary
search algorithm was used to correct if any de-skewing in the image. Dot parts
were detected from the image and equivalent braille character was recognized
using matching algorithm. The methodology used in this paper was experimented
on Thirukkural Braille Book and achieved a result of more than 90% accuracy.
Padmavati et.al 6 projected a research to convert
Braille script document into its corresponding letters of 3 languages i.e
English, Tamil and Hindi. Pre- processing techniques like Gaussian filter were performed
on the Braille document to improve the dots and to eliminate the noise. Piece
wise enhancement techniques such as contrast stretching, intensity stretching was
used for enhancing the dots. As a next step, the edge detectors and projection
profile method were applied to crop the interest area. The image was first separated
into lines and then into Braille cells by applying horizontal and vertical
projection profiles. A Binary pattern vector of length 6 for each Braille cell
was generated. Binary 1 and 0 were used to represent presence and absence of braille
dot respectively. The corresponding alphabet were generated using its pre-built
Srinath et.al 7
presented an Optical Braille character recognition system for Kannada Braille document.
The project took the image of Kannada Braille script and segmented the image in
to line by using the relative position of the dots. After line segmentation was done, Braille
characters were separated using inter character distance parameter. The
recognized Braille character was translated into Kannada alphabet and saved in
a document. The methodology used in this paper achieved more than 98% accuracy.
Ravi et.al 8 projected
a research to convert hand punched Kannada Braille Characters using knowledge
based multi decision method. Braille dots were carefully separated depending on
the location of the dots. The inter character distance was used to group the
dots into a word box. The system was designed to recognize the braille dot and
it was converted to Kannada character.
9 presented a research work to convert Odia, Hindi, Telugu
and English braille documents into its corresponding language.
The algorithm used the
technique of histogram analysis, segmentation, pattern recognition, letter
arrays, data base generation with testing in software and dumping in using
Spartan 3e FPGA kit which defined the dot patterns for the alphabets.
Rao et al. 10 proposed
a research work to convert Kannada Braille document
taken by a camera, into Kannada script or audio. As a first step the color
image of input image was converted to unicolor space for processing. An automated thresholding algorithm was applied to
get the area of interest and segmentation technique was applied and recognition
of letter was done based on highlighted dot in Braille document. All algorithms
were implemented for a Xilinx Spartan 3E FPGA using Verilog HDL language
and were executed in real time. An accuracy of
over 94% was attained in Braille segmentation and detection.
Ann Jose et.al 11 projected
a work on changing the Malayalam Braille document to text and concatenative
speech synthesis technique was used for speech conversion. The image was
captured by CMOS image sensor and then the image was converted to binary image
by calculating the threshold using histogram analysis. Filtering technique was
applied to remove the noise. The row and column grouping of dots was done based
on the spacing between dots to identify the Braille cell. Presence of dot will
be represented as 1 and absence as 0 and a six-digit binary number was generated.
Using this binary number Malayalam character mapping was done and speech synthesizer
was used for audio conversion.
Conclusion & Future
has been developed as the reading and writing system for the visually
challenged people. Few methods have been proposed in the past for recognition
of Braille script in Southern Indian Languages based on histogram technique,
piece wise enhancement techniques, image filtering, horizontal profiling,
vertical profiling, image thresholding, de-skewing the image, template matching
etc. These approaches considered the different attributes related to scanned
document orientation, alignment, contrast, color, intensity,
connected-components, edges etc. These attributes are used to classify dot
regions in the image. This paper provides a study of the Braille recognition in Southern Indian
Languages proposed earlier. Every method has its own benefits and limitations. The
future effort mainly focusses on developing a single unified approach for efficient
and better braille document recognition in
Southern Indian Languages.