A MOBILE APPLICATION FOR HANDWRITING RECOGNITION USING MACHINE LEARNING AND IMAGE PROCESSING

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 10 Issue: 09 | Sep 2023 www.irjet.net p-ISSN: 2395-0072
© 2023, IRJET | Impact Factor value: 8.226 | ISO 9001:2008 Certified Journal | Page 188
A MOBILE APPLICATION FOR HANDWRITING RECOGNITION USING
MACHINE LEARNING AND IMAGE PROCESSING
SATHEESH N P 1, SURYA S S2, SAMYUKTHA K R P3
1 Professor, Dept. of Artificial Intelligence & Data Science, Bannari Amman Institute of Technology, Tamil Nadu, India
2 Student, Dept. of Information Technology, Bannari Amman Institute of Technology, Tamil Nadu, India
3 Student, Dept. of Information Technology, Bannari Amman Institute of Technology, Tamil Nadu, India
---------------------------------------------------------------------***-----------------------------------------------------------------
Abstract—This paper revolves around character recognition,
with particular focus on its application to handwriting
education. The current handwriting teaching method
landscape faces a major challenge of lacking timely and
personalized feedback to learners and individual care.
Traditional methods rely heavily on human evaluation which
can be slow and subjective. Therefore, there is an urgent need
for innovative solutions that use technology to fill this gap. The
main purpose of this paper is to develop a real-time character
recognition system that can provide feedback to users, thereby
improving the handwriting teaching process. A central problem
is to create an efficient character recognition system that
seamlessly integrates machine learning and image processing
techniques. To achieve this, an extensive dataset of handwritten
characters is carefully collected to ensure diversity of writing
styles and variations. This dataset serves as the basis for
training a machine learning model. Our results show impressive
accuracy, with an average recognition rate of over 95% for a
wide range of handwriting styles and their variations. In the
discussion that follows, we interpret these findings and
highlight the significant impact of our system in improving the
learning experience. The real-time feedback mechanism
introduced by our solution streamlines the teaching process
and encourages students to significantly improve their
handwriting skills. In summary, this study effectively addresses
the urgent need for effective handwriting teaching tools
through the synergy of machine learning and image processing
techniques.
Key Words: Character recognition, handwriting, machine
learning, image processing, education, personalized feedback.
1. INTRODUCTION
Education plays a crucial role in ensuring human survival and
well-being. It empowers individuals with knowledge,
strengthens their character, broadens their perspective on life,
instills ideals, and equips them with the ability to adapt to
evolving environments. Consequently, every citizen is entitled
to a high-quality education [1]. Regrettably, the UIS Global
data for the 2018 academic year reveals a distressing reality:
more than 59 million primary school children remain
deprived of educational opportunities [2]. This issue is
particularly acute in rural areas, where many children resort
to selling goods on the streets or laboring in fields to
supplement their family's income, as detailed in [3].
Compounding this problem, rural schools are often sparse,
forcing students to endure arduous daily journeys of over
three kilometers to reach school. Furthermore, the shortage of
resources and teachers poses significant barriers to providing
quality education universally. The question then arises: How
can we ensure equal access to quality education for all
children? To address this pressing challenge, we propose the
development of an Android application tailored for primary
education.
2. LITERATURE SURVEY
Nazmus Saqib et al.,[4] (2022) This study explores how
handwriting recognition technology has been adopted in the
industry, but not great, impacting both performance and
usability. Therefore, the character recognition technology
used is not yet very reliable and needs further improvement
to be widely used for serious and reliable tasks. In this his
account, recognition of English alphabet letters and numbers
is performed by proposing a custom-made his CNN model
using his two different datasets of handwritten images, Kaggle
and MNIST respectively. will be These models are lighter, yet
offer greater accuracy than the latest models.
Asif Karim et al.,[5] (2021) In this study, we demonstrated the
feasibility of recognizing handwritten images from a given
input data set (MNIST) using a convolutional neural network
(CNN). The researchers curated his MNIST dataset, which
contains over 60,000 images, and trained a CNN model to
classify characters with a high accuracy of 99.70% while
testing the model on approximately 10,000 image samples.
Hemangee Sonara and Dr. Gayatri S Pandi [6] (2021)
proposed a paper using convolutional neural network (CNN)
algorithm for recognizing handwritten characters. First, the
input image is denoised using a median filter to segment the
image. Next, perform feature extraction and recognition from

the input image. The system should provide users with better
quality of service and higher character recognition accuracy.
3. OBJECTIVE AND METHODOLOGY
3.1 OBJECTIVE 1-HANDWRITTING RECOGNITION
AND ANALYSIS
One of the main objective of the paper revolves around
developing a machine learning model that can recognize and
analyze handwritten characters and words. This goal can be
broken down into more important sub-objective, each of
which contributes significantly to the achievement of the
overall objective.
➢ Advanced image preprocessing: The goal is to implement
advanced image preprocessing techniques to improve the
quality of input handwritten images. This can be achieved by
various techniques includes noise reduction, contrast
enhancement, and image format standardization.
➢ Model training for high accuracy: The goal is to train a
machine learning model using a diverse dataset of
handwritten patterns. The focus is on achieving high accuracy
in character and word recognition so that the application can
provide accurate feedback to the user.
3.1.1 INTERACTIVE LEARNING INTERFACE
The second objective focuses on designing an interactive and
user-friendly mobile application interface for effective
handwriting lessons.
➢ Responsive User Interface Design: The goal is to create
engaging, userfriendly user interfaces that promote a positive
learning experience. The user interface is designed with user
engagement and learning outcomes in
mind. The user interface must be simple and neat so that it can
fit to any users without age limit and it must be easily
operatable so that it cannot cause any problems for users.
➢ Real-time feedback: Seamless integration of handwriting
recognition models into applications is critical. The objective
is to give users real-time feedback as they practice their
handwriting, dynamically improving their skills.
3.2 FLOW DIAGRAM
A visual representation of our paper's workflow is
encapsulated in the following flow diagram:
EXPLANATION OF FLOW DIAGRAM:
• Data Collection: In the first stage, a diverse dataset of
handwriting samples is collected. This dataset serves as the
basis for training and validating machine learning models.
• Preprocessing: The raw input image undergoes a series of
preprocessing steps to remove noise, enhance contrast, and
standardize format. This step ensures the high quality and
consistency of the input data.
• Feature Extraction: Relevant features are extracted from
pre-processed handwritten images. These features capture
important handwriting features required for recognition and
analysis.
• Machine Learning Model: A machine learning model is
developed and trained using the extracted features and the
tagged dataset. This model forms the core of our handwriting
recognition system.

• Handwriting Recognition: Trained machine learning models
are checked with some real time input images and the
accuracy is checked and improved the accuracy of the model
by training rigorously with multiple test images.
• User Interface Design: In parallel, we design an interactive
and user-friendly interface for the mobile application. This
interface plays a crucial role in engaging users and enhancing
their learning experience.
• Integration: Trained machine learning models are integrated
into mobile applications to enable real-time handwriting
recognition. Users can get instant feedback on their
handwriting practice.
• User Personalization: This is where our machine learning
algorithms come into play, adjusting materials and exercises
to individual progress. This application provides customized
recommendations to improve each user's handwriting skills.
• Dynamic Learning Environment: An integrated system
creates a dynamic learning environment. This environment
encourages user engagement and continuous improvement of
handwriting skills. Using tablets include the cost of tablets, the
need for technical support, and potential distractions.
4.HANDWRITTING RECOGNITION
During the initial development phase, machine learning
played a crucial role in investigating the most effective method
for comparing handwriting. This phase also involved
discovering the optimal parameters for image processing. To
facilitate this process, an application for recording characters
was created and utilized to gather handwriting samples from
teachers. These teacher-written characters were subjected to
the same testing and served as models to instruct children on
how to write these characters. Drawing upon the insights
gained from these two phases, an application was designed,
developed, and subjected to rigorous testing in the final phase.
Figure 2 provides a visual representation of these distinct
phases.
4.1 DATASET
The embark on a detailed exploration of the key components
comprising the Dataset required for the paper module. The
dataset that needs to be choose must be a larger in size and
must contain handwritten characters and digits images.
Initial Phase: The initial stage of paper begins with a
comprehensive look at the development stages of the entire
paper. This is a preliminary stage where machine learning
emerges as a guiding light. This phase represents the
emergent phase of the paper, and careful planning is critical.
Machine learning plays a central role in finding the best
approach to character comparison and optimizing the best
parameters for image processing. The importance of this
phase cannot be overemphasized. It forms the basis of the
entire paper. In order to build a machine learning model that
can seamlessly recognize the handwriting from the images,
the first step in building a model will be choosing the
appropriate dataset. The dataset that needs to choose must be
large in size and must contain a variety of handwritten images
so that machine learning model can train and perform very
well.

IAM Dataset Testing: The paper study continues with
a key point in the development of handwriting recognition
systems: The testing phase. Here we searched deeper into the
area of datasets, specifically IAM Dataset (an English sentence
database for offline handwriting recognition). This dataset
serves as a testing ground, a crucible, for testing and refining
the effectiveness of various classifiers. In this vast
environment, we examined three classifiers: Decision Trees,
Naive Bayes, and K-Nearest Neighbors. These classifiers are
faithful companions on this journey when it comes to
evaluating their abilities and measuring their ability to achieve
accurate handwriting recognition. We scrutinize key
performance indicators, analyze successes and failures, and
constantly seek the elusive goal of error-free handwriting
recognition.
4.2 CNN AND TENSORFLOW
The implementation of Convolutional Neural Networks (CNN)
using TensorFlow, a critical aspect of our Machine Learning
and Handwriting recognition module. CNNs have proven to be
a powerful tool in image classification tasks, making them an
ideal choice for handwriting recognition within our
application.
CNN Architecture: We begin by discussing the
architecture of our CNN model. This includes the number of
convolutional layers, pooling layers, and fully connected
layers. Each layer's purpose in the network is explained,
providing insight into the decision-making process behind our
architecture. The CNN model is implemented using
TensorFlow. The CNN model must perform well so that
various factors (Convolutional layers, Pooling layers, Fully
connected layers) are carefully chosen in order to achieve the
best performing CNN model.
Data Preparation: The preparation of data for training
and testing is a crucial step in building a robust handwriting
recognition system. Here, how the dataset will be
preprocessed and augmented to ensure that our CNN can
learn effectively from the provided data. Data augmentation
techniques, such as rotation, scaling, and noise addition, are
discussed in detail for gaining more accuracy of the CNN
model. The split ratio of the dataset (for e.g., 80:20) so that the
model can train and perform accurately.
Training Process: The paper work continues with an
exploration of the training process. We outline the parameters
used during training, including learning rate, batch size, and
the number of training epochs. The convergence of the model
and the adjustments made during training are thoroughly
examined. The number of training epochs is planned
minimum of 150 so that the predication accuracy of the model
can be more because of the number of training epochs.
5. MOBILE APPLICATION DEVELOPMENT
In this mobile app, users begin their journey by signing in with
their Gmail credentials via the “Sign in with Google” option.
Once authenticated, they will be seamlessly transferred to the
main interface of the app, where an unmistakable pencil icon
located in the bottom right beckons them to access the image
upload function. On the image upload screen, the user has the
choice to submit a handwriting template, which can be
manually selected from the device's photo library or taken
fresh with the device's camera.
After selecting the desired image, simply pressing the "Check
my handwriting" button triggers the transmission of the
selected image to the application's backend server for careful
processing. These backend systems use trained ml model to
review submitted handwriting, and once the analysis is
complete, the app will proudly display the user's handwriting
skill score prominently. on the main screen. Additionally, to
ensure quick delivery of results, scores are sent
simultaneously as push notifications, ensuring users have
quick access to their handwriting.
For those concerned about security and data storage, the app
goes a step further by automatically deleting all data stored
locally on the user's mobile device when they choose to sign
out. This comprehensive process is meticulously designed to
provide users with an accessible and insightful handwriting
analysis experience that seamlessly combines convenience
and information. This whole process is shown in the form of
flow chart (Figure 3).
5.1 HANDWRITING RECORDING
The accurate detection of a character does not necessarily
guarantee that it adheres to correct handwriting standards or
aligns with the way children learn to write. In fact, the
handwriting samples found in the IAM database are not
directly comparable to the characters typically found in
children's learning materials. Consequently, while the findings
from the previous analysis provide valuable guidance, they do
not prescribe a specific classifier or algorithm to be employed
in the educational application.

Figure 3: Flow chart for the mobile app
5.1.1 The Recorder App:
To be able to have handwriting more comparable to the
handwritten characters in the final application, a “recorder
app” has been developed. This app will have the main screen
where the user will have to select the pencil symbol at right
bottom corner (Figure 4).
Figure 4: Main Screen
On the image upload screen, the user has the choice to submit
a handwriting template, which can be manually selected from
the device's photo library or taken fresh with the device's
camera. After selecting the desired image, simply pressing the
"Check my handwriting" button triggers the transmission of
the selected image to the application's backend server for
careful processing. These backend systems use the trained ml
model to review submitted handwriting, and once the analysis
is complete, the app will proudly display the user's
handwriting skill score prominently. on the main screen.
Additionally, to ensure quick delivery of results, scores are
sent simultaneously as push notifications, ensuring users have
quick access to their assessment.
5.1.2 Results
Classifier Subset of
characters
J48: Decision Tree 41.6%
Naïve Bayes: Bayesian Classifier 72.9%
Lazy IBK: Instance Based Leaner 65.8%
Table 1: Result for 10-fold validation

Regrettably, the outcomes were less favorable when applying
the same classifiers to a subset of the dataset consisting of
characters recorded by teachers (as shown in Table 1).
Despite achieving the highest success rate of nearly 75% on
characters that, in theory, should closely resemble one
another (as per the teachers' instructional goals for children),
this approach does not appear promising. As a result, an
alternative method must be sought.
5.2USER INTERFACE
Our research begins with a detailed analysis of the mobile
application itself. Mobile applications are innovative tools
designed to bridge the gap between traditional handwriting
and the digital realm. We will delve into the architecture of the
app and discuss its design principles and functionality. The
app has been carefully designed so that the characters
generated are very similar to those found in educational books
for children. We analyzed the intuitive interface to seamlessly
integrate the user into the character drawing process. We
break down the drawing process that is at the heart of the app
and help readers understand the methodology behind
handwriting generation. Additionally, we are introducing the
app's Data Export feature, a core feature that streamlines the
process of collecting and organizing handwriting for reference
and study.
5.3 INTERACTIVE FEATURES
Results: Our search has progressed to a point where
we focus on results. Here we present the results of the test
phase, in which the letters recorded by the teacher are
rigorously examined. These results provide valuable insight
into the effectiveness of our unique approach. However, it also
highlights limitations and challenges associated with this
methodology. Recognizing these limitations acts as a catalyst
for innovation and drives us to explore alternative methods.
Alternative Method: This method is about creating
and evaluating baselines. This methodology is both effective
and complex. We undertake a comprehensive investigation of
this alternative approach, analyzing its components and
uncovering its inner workings. Every aspect is carefully
examined, from baseline creation to the complexity of the
scoring algorithm. The success rate of this alternative method
is a staggering 90%, demonstrating a detailed understanding
of how this result is achieved.
5.4 APPLICATION DEVELOPMENT
The application development module that is the main feature
of this paper. This includes integration, testing, and creating
user-friendly interfaces. This chapter describes goals,
methods, and insights related to this important stage.
App Development: The application is created as per
the UI design made for the uploading area and that it shows
the feedback immediately after user uploads the handwritten
image either from device or can take a picture from mobile
camera. The application’s front end is developed using Kotlin
that it gets the image from the user and send it to backend
using JSON.
Figure 5: User Dashboard
Additional Features: In addition, the main screen also
acts as a user dashboard which saves all previous uploads
from the user and the marks and the received personalized
feedback for his/her handwriting image that had been
uploaded in the mobile app (Figure 5).

5.5 APPLICATION INTEGRATION WITH TRAINED ML
MODEL
Trained ML Model Integration: Here, we detail the
integration of a trained Machine Learning model into the
application. This model serves as the backbone for
handwriting recognition and scoring, enhancing the
educational experience. We elaborate on the selection of ML
algorithms, the training process, and the deployment within
the app. The trained ML model is saved using TensorFlow
library and integrated into the mobile application.
Image Processing Optimization: The heart of image
processing lies in precision and efficiency. We delve into the
selection of libraries and tools, with a focus on OpenCV. The
journey from raw images to processed data becomes an art,
and we explore how this artistry is achieved.
Testing: The application is now tested with the
integrated ML model. The performance of the mobile app and
its stability is closely monitored. The image recognition model
performance also noted and its accuracy is calculated. The app
is tested severely with various user in order improve its
accuracy and various bugs can be identified and can be
corrected.
These chapter provide a comprehensive understanding of
modules, methodologies, and the integration of ML models
driving the development of the Handwriting Teaching Mobile
Application. From handwriting recognition to user interface
design, and application development to testing, innovation
and optimization have enriched our paper. This lays a robust
foundation for an educational tool set to revolutionize
handwriting education through cutting-edge technology.
6. IMAGE PROCESSING
Regrettably, the outcomes were less favorable when applying
the same classifiers to a subset of the dataset consisting of
characters recorded by teachers (as shown in Table 1).
Despite achieving the highest success rate of nearly 75% on
characters that, in theory, should closely resemble one
another (as per the teachers' instructional goals for children),
this approach does not appear promising. As a result, an
alternative method must be sought.
We conducted a comparison between Java (Android) libraries,
including but not restricted to JMagick and OpenCV, to
determine which library could perform this task with the least
processing time. Ultimately, OpenCV was chosen for
implementation in the app to manage image processing.
However, it's worth noting that using OpenCV necessitates an
additional app on the device for it to function optimally. This
requirement is in place to ensure the app can deliver the
highest level of performance and speed, as it is specifically
optimized and compiled to leverage the device's core
capabilities.
6.1 IMAGE DETECTION AND PREPROCESSING
Image Detection: The implementation journey
commences with image detection, a critical step in the process
of recognizing handwritten characters. We explore various
techniques for identifying characters within images, including
object detection algorithms.
Object Detection Algorithms: While experimenting
with different methods, including OpenCV's Haar cascades, we
ultimately opt for the histogram of oriented gradients (HoG)
descriptors in combination with Support Vector Machines
(SVM) for object detection. This decision is made based on the
superior performance of the HoG-SVM approach in
recognizing characters.
Image Preprocessing: Once characters are detected,
preprocessing steps are crucial to enhance the quality of the
images. We discuss procedures such as grayscale conversion,
consistent resizing, and noise removal. Additionally, we
employ a Gaussian filter to further refine the images.
6.2 MACHINE LEARNING INTEGRATION
Support Vector Machines (SVM): To classify the
extracted features and recognize characters, we employ
Support Vector Machines (SVM), a popular and robust
machine learning algorithm. SVM is well-suited for multi-class
classification problems, making it an ideal choice for
handwriting recognition in our paper.
Classification and Comparison: SVM is employed in
conjunction with the recovered facial attributes to distinguish
between various character classes. We discuss how SVM's
effectiveness is assessed by comparing its performance with
other algorithms like logistic regression and random forest.
6.4 TRAINING AND TESTING
Training and Testing Database: One of the core
challenges in machine learning is the development of effective
algorithms. We detail the processes of data collection and
dataset creation, which includes labeling and feature
extraction. This dataset serves as the foundation for training
and testing our machine learning models.

Cross-Validation: Cross-validation techniques are
used to ensure unbiased model evaluation. We describe the
division of the dataset into training and testing sets, with
considerations for different split ratios (e.g., 70:30 and 80:20).
Multiple cross-validation levels, including 4, 5, and 10, are
employed to assess model performance comprehensively.
This chapter provides insight into the application of machine
learning and image processing techniques to achieve
handwriting recognition and improve the learning experience
with the help of the mobile application. The emphasis is on
accuracy, adaptability and continuous improvement to ensure
the effectiveness of the application in teaching and improving
handwriting skills.
7.RESULT AND DISCUSSION
This section presents the results in a structured way according
to the methodology used throughout the paper. This includes
compilation of results in the form of images, graphics and
tables, as well as detailed descriptions of key findings.
7.1.1 Machine Learning Model Performance
This paper introduced a handwriting recognition system
based on convolutional neural networks (CNN). The objective
is to enhance handwriting instruction by providing students
with interactive feedback based on the accuracy of their
writing. The dataset used in the study comprises 55,000
photographs of handwritten images, which have undergone
preprocessing techniques including grayscale conversion,
normalization, and resizing to 28x28 pixels. The CNN model
consists of two convolutional layers, two fully connected
layers, and a SoftMax layer for classification. Model training is
achieved through backpropagation and the Adam optimizer,
utilizing the preprocessed dataset. The system was subjected
to testing using a set of 10,000 photos, achieving an
impressive accuracy rate of 98.4%. Furthermore, the study
conducted a comparative analysis with other state-of-the-art
handwritten recognition systems, demonstrating its
superiority in terms of both accuracy and computational
efficiency. The practical implementation of this system in a
handwriting class holds significant promise, as it can provide
students with immediate feedback on the accuracy of their
writing and tailor assignments to their individual writing
skills. Consequently, this study underscores the advantages of
employing CNN algorithms for handwriting recognition and
highlights the practicality of applying this technology to the
teaching of handwriting.
7.1.2 Mobile Application User Engagement
User engagement was assessed by analyzing user interactions
with the mobile application. Table 2 presents the scores of 6
participants used the app. The second column shows the time
take to upload the user image i.e., the handwritten image,
while the 3rd column represents the time take to process the
user given image and the 4th column shows the time taken by
the ML model to predict the mark for the handwriting of the
user from the user given image and the last column shows the
mark given by the model for the user.
Table 2 – Test Results
The comparison of score indicates the marks is predicted from
the training model is comparably accurate and the feedback is
provided based on the score of the user handwriting score.
The z score for the above results is 1.8
7.1.3 Various Test Results
The model processes the request send from various users in
the FIFO (First In First Out) order. It correctly predicts the
score based on the handwritten image send from the user. The
model that handles the multiple request (Figure 8) clears
shows the model processes the image and send the marks for
the processes image.
Student Image
Upload
Time
Image
Process
Time
Mark
Prediction
Time
Marks
Given
1 3s 2.6s 2s 9
2 2s 2.4s 1.9s 7
3 2.5s 2.4s 1.7s 4
4 2.7s 2.3s 1.8s 1
5 2.9s 2.8s 1.8s 7
6 2.6s 2.5s 1.9s 8
Average 2.6s 2.5s 1.9s 6.6

Figure 8: ML model handles multiple request
7.2 HANDWRITING RECOGNITION ACCURACY
This section details the results, from simple to complex,
according to the numbers given. This discussion includes the
significance of the results, comparison with relevant
publications, and their impact. This result includes the
machine learning model performance, handwriting
improvement, comparisons with the related studies.
7.2.1 Machine Learning Model Performance
The results show that the machine learning model performs
very well in handwriting recognition with an accuracy of over
90%. The Confusion Matrix (Figure 9) provides insight into
where the model does well and where it needs improvement.
Figure 9: Confusion Matrix
7.2.2 Handwriting Improvement
An observed over 80% improvement in the readability of
students' handwriting after using our application is a positive
result. Table 3 visually confirms the significant improvement
in handwriting quality, highlighting the educational value of
the application.
Table 3- Legible Handwriting
7.3 SIGNIFICANCE, STRENGTHS, AND LIMITATIONS
7.3.1 Significance
Our paper holds immense significance in the field of
handwriting education. By providing a technological solution
that enhances legibility and engages students, we contribute
to modernizing the teaching methods in this domain.
7.3.2 Strengths
Our paper's strengths lie in its user-friendly interface,
effective handwriting recognition, and proven improvements
in handwriting quality. The synergy of Image Processing and
Machine Learning has resulted in a robust and innovative
educational tool. This paper also provides immediate feedback
on user handwriting and the score for user handwriting that is
a major strength of this mobile application.
Skills SD
1. Properly shaping letters .93 3.43
2. Size of Letters .79 3.79
3. Slope 1.03 3.11
4. Connection .78 3.68
5. Extensions .76 3.89
6. Line .75 3.62
7. Space b/w words .74 3.71
8. Cleanliness .89 4.29
9. Page format .97 4.21
10. Perfect Writing .81 3.61
Total .71 3.91

7.3.3 Limitations
Despite its strengths, our paper has limitations, such
as the need for continuous updates to accommodate diverse
handwriting styles. Additionally, accessibility challenges may
arise for students without access to mobile devices or the
internet. The app always requires internet which will be one
of the major limitations of this mobile application.
This chapter contains a detailed presentation of the paper
results and findings. We highlight the importance of these
findings, compare them with related studies, and identify the
strengths and limitations of our paper. Taken together, these
results highlight the value and impact of machine learning and
image processing for handwriting education mobile
applications in the field of handwriting education.
8.CONCLUSION AND SUGGESTION FOR FUTURE
WORK
The preliminary results from tests conducted on a small group
of students have demonstrated promising outcomes.
However, we acknowledge that effective handwriting
instruction involves more than character recognition; factors
such as starting points and movement directions play critical
roles. Therefore, future studies will delve deeper into
understanding the impact of our approach on the learning
curve, student motivation, and teacher engagement.
SUGGESTION FOR FUTURE WORK
As we conclude this paper, we recognize that certain
aspects remain unexplored and hold the potential for further
refinement and expansion. Here are key areas for future
research and development:
• Enhanced Character Recognition: While our paper
achieved a commendable recognition accuracy, there is room
for improving the model's performance, especially when
dealing with diverse handwriting styles and languages and it
can be used to recognize specific characters and that can be
used to teach handwriting in a more efficient manner.
• Teacher-Student Collaboration: Exploring ways to
enhance collaboration between teachers and students within
the application, facilitating seamless feedback and progress
tracking.
• Integration of More Writing Elements: Expanding the
application's scope to cover cursive writing, different
languages, and more complex writing elements to address a
broader audience.
• Gamification and Engagement Strategies:
Investigating the incorporation of gamification elements and
engagement strategies to make the learning process more
enjoyable and effective.
• Deployment in Educational Sectors: Collaborating
with educational institutions to deploy the application in real
classrooms, gathering valuable insights for further
improvement.
In conclusion, while our paper has laid a strong foundation for
improving handwriting education through technology, there
are exciting possibilities for future research and development,
ensuring a more effective, engaging, and inclusive learning
experience for students of all ages and backgrounds.
REFERENCES
[1] Sudarsana, K., Nakayanti, A.R., Sapta, A., Haimah, Satria, E.,
Saddhono, K., Daengs GS, A., Putut, E., Helda, T., Mursalin, M.
(2019). Technology Application in Education and Learning
Process. J. Phys. Conf. Ser.
https://www.researchgate.net/publication/337249285_Tech
nology_Application_In_Education_And_Learning_Process.
[2] Education: Out-of-School Rate for Children of Primary
School Age. (2022).
https://data.worldbank.org/indicator/SE.PRM.UNER.ZS
[3] Weinstein, J. The Problem of Rural Education in the
Philippines. (2010).
https://joshweinstein.wordpress.com/2010/03/02/the-
problem-of-education-in-the-philippines/
[4] Butler, C., de Pimenta, R., Fuchs, C., Tommerdahi, J.,
Tamplain, P. (2019). Using a handwriting app leads to
improvement in manual dexterity in kindergarten children.
Res. Learn. Technol.
https://journal.alt.ac.uk/index.php/rlt/article/view/2135
[5] Pegrum, M., Oakley, G., Faulkner, R. (2013). Schools going
mobile: A study of the adoption of mobile handheld
technologies in Western Australian independent schools.
Australas. J. Educ. Technol.
https://ajet.org.au/index.php/AJET/article/view/64

[6] Stephanie, Handwriting Apps for Kids (2020).
https://parentingchaos.com/handwriting-apps-for-kids/
[7] Baldominos, A., Saez, Y., Isasi, P. (2019). A survey of
handwritten character recognition with mnist and emnist.
Appl. Sci.
https://www.mdpi.com/2076-3417/9/15/3169
[8] Dewa, C.K., Fadhilah, A.L., Afiahayati A.(2019).
Convolutional neural networks for handwritten Javanese
character recognition. Indones. J. Comput. Cybern. Syst.
https://www.researchgate.net/publication/322892771_Conv
olutional_Neural_Networks_for_Handwritten_Javanese_Chara
cter_Recognition
[9] Lincy, R.B., Gayathri, R. (2021).Optimally configured
convolutional neural network for Tamil Handwritten
Character Recognition by improved lion optimization model.
Multimed. Tools Appl.
https://www.researchgate.net/publication/344612091_Opti
mally_configured_convolutional_neural_network_for_Tamil_H
andwritten_Character_Recognition_by_improved_lion_optimi
zation_model
[10] Zhang, Z., Tang, Z., Wang, Y., Zhang, Z., Zhan, C., Zha, Z.,
Wang, M. (2021).Dense Residual Network: Enhancing global
dense feature flow for character recognition. Neural Netw.
https://www.sciencedirect.com/science/article/pii/S089360
8021000472
[11] Nazmus Saqib, Khandaker Foysal Haque, Venkata
Prasanth Yanambaka, Ahmed Abdelgawad.(2022).
Convolutional-neural-network-based handwritten character
recognition: an approach with massive multisource data.
https://www.mdpi.com/1999-4893/15/4/129
[12] Asif Karim, Pronab Ghosh, Atqiya Abida Anjum, Masum
Shah Junayed.( 2021). A Comparative Study of Different Deep
Learning Model for Recognition of Handwriting Digits.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=37692
31
[13] Hemangee Sonara, Dr. Gayatri S Pandi. (2021).
Handwritten Character Recognition using Convolution Neural
Network.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=39220
13
[14] Musbau Dogo Abdulrahaman, Nasir Faruk, Abdulkarim
Oloyede, Nazmat Surajideen- Bakinde. (2020). Multimedia
tools in the teaching and learning processes: A systematic
review.
https://pubmed.ncbi.nlm.nih.gov/33195834/

A MOBILE APPLICATION FOR HANDWRITING RECOGNITION USING MACHINE LEARNING AND IMAGE PROCESSING

More Related Content

Similar to A MOBILE APPLICATION FOR HANDWRITING RECOGNITION USING MACHINE LEARNING AND IMAGE PROCESSING (20)

More from IRJET Journal (20)

Recently uploaded (20)

A MOBILE APPLICATION FOR HANDWRITING RECOGNITION USING MACHINE LEARNING AND IMAGE PROCESSING