Vol. X, Issue 1 (June 2024) [Link]

in/

Human Computer Interaction based Head Controlled

Mouse
Vadde Veeresh*
Pothuraju Ajay, Nandimandalam Bharat Varma†
Jukanti Anusha, Ladineni Divija‡

Abstract
It's incredibly difficult for someone with physical disabilities to control the
mouse. We have suggested utilizing the head to control the mouse pointer in
order to provide a solution for those who are unable to use a real mouse. This
is a different approach to utilizing your head and eyes to manipulate the
mouse when accessing a computer. One essential real-time input method for
human-computer communication is head movement, which is crucial for those
with physical disabilities. A revolutionary eye control method is provided in
this system that uses a webcam and doesn't require any additional hardware
in order to increase the eye tracking technique's mobility, usability, and
reliability in user-computer communication. The goal of the suggested system
is to offer an easy-to-use interactive mode that solely makes use of the user's
facial movements. The suggested solution explains how to utilize a webcam
and Python to control the cursor on the screen by implementing both iris and
head position tracking. It's incredibly difficult for someone with physical
disabilities to control the mouse. We have suggested utilizing the head to
control the mouse pointer in order to provide a solution for those who are
unable to use a real mouse. This is a different approach to utilizing your head
and eyes to manipulate the mouse when accessing a computer. In this system,
a unique eye control system is proposed to enhance the eye tracking
technique's usability, mobility, and reliability in user-computer
communication. It does this by using a Webcam and doesn't require any
additional hardware. The goal of the suggested system is to offer an easy-to-
use interactive mode that solely makes use of the user's facial movements. The
suggested method uses machine learning with Python to implement both iris
and head position-based cursor movement, which may be utilized to control
the cursor on the screen via a webcam.

*Associate Professor, Dept of CSE, ACE Engineering College, Hyderabad, India. Email:
vaddeveeresh@[Link]
†Student, Dept. of CSE, ACE Engineering College, Hyderabad, India.
‡Student, Dept. of CSE, ACE Engineering College, Hyderabad, India.

MZUJHSS, Vol. X, Issue 1, June 2024 268

 Keywords: Head Controlled Mouse, Gesture Recognition, Human Computer
Interaction, User Interface Design.

Introduction
These days, a lot of people suffer from conditions like paraplegia, which physically
prevents a person from utilizing their body below the neck. For this reason, a face gesture
control system is required so that users with physical limitations can operate a computer using
facial gestures, such as eye movements for cursor movement and eye winking for clicking.
Input devices like a mouse, keyboard, scanner, etc. are typically used to communicate with
electronic devices like computers. This makes it challenging for those who have physical
limitations to communicate. Cursor control using face gestures is a cutting-edge technology
that enables users to manipulate the movement and position of a computer cursor solely
through facial expressions and gestures. By eliminating the need for conventional input
devices like a mouse or trackpad, this technology offers a more intuitive and hands-free
interaction experience. This technology holds tremendous potential for various applications,
including accessibility for individuals with limited mobility, virtual reality experiences, and
hands-free computing in environments where traditional input devices are impractical or
inconvenient. With cursor control using face gestures, users can perform a range of actions
such as moving the cursor, clicking, scrolling, and even executing specific commands by
mapping facial gestures to predefined functions. Common facial gestures used in this
technology include raising eyebrows, winking, smiling, nodding, or tilting the head. The
system accurately recognizes and translates these gestures into corresponding cursor
movements or actions on the screen. This innovative approach to cursor control not only
provides a novel and engaging user experience

Objectives
In our paper there are 5 objectives. They can be listed as:
a) Face and Eyes Detection.
b) Extraction of Eye Corners
c) Create an algorithm to determine the point of gaze based on the identified eye
attributes.
d) Control the mouse clicks with the eye blinks.
e) To control the cursor with head movement

Methodology
Review inputs and outputs for project activities. Information will be collected and
prioritized. An appropriate algorithm or framework has been selected. Several estimation
algorithms will be compared and the best method will be selected. Software and hardware
selection will be made according to the needs. Data will be used as a process or framework

MZUJHSS, Vol. X, Issue 1, June 2024 269

 Fig. 1: System Architecture

Literature Survey
Although Drewes [1], Heiko [2] (2009) provides a thorough description, it was
highlighted that because most algorithms required laborious and time-consuming techniques
to calibration, they needed to be improved. Heiko Drewes and Albrecht Schmid [3]. In 2017
an update was released. This method was created for patients who are paralyzed. This method
uses a person's pupil to control the mouse cursor via MATLAB. Karlene Nguyen [4], Cindy
Wagner [5], David Koons [6], Myron Flickner [7]. A wearable eye-gaze monitoring system
based on vision was introduced in 2016. A high-infrared camera is used by this system to
operate. With the use of the infrared camera, it can identify the subject's eyes. The problem
with this approach is that it is expensive and cumbersome. Aleksei Bukhalov [8], Viktoria
Chafonova [9]. The Pupil Center Manages Detection 2015 saw the introduction of the
Circular Hough Transform Technique. The suggested application's algorithm uses the dark-
pupil technique to identify the pupil edge circle from a set of infrared camera-taken eye
pictures. Cheng-Chih Wu [10] 2014 saw the creation of a MATLAB-based face and eye-
controlled system. Using a camera, the mouse may be moved by adjusting the face and eyes.
This technology's limitation is that it can only operate a few centimeters away from the
source.

Proposed System
With the help of gestures and facial expressions, users will be able to manipulate and
control the cursor on a computer screen with the help of the suggested face gesture cursor
control system. Using computer vision and face recognition algorithms, the system tracks and
interprets particular facial movements in real time, converting them into exact commands or
cursor motions. Typically, the essential elements of the suggested system consist of:
Recognition of Facial Gestures: The system uses models and algorithms to precisely identify
and track facial gestures such head tilts, grins, blinks, and brow movements.

MZUJHSS, Vol. X, Issue 1, June 2024 270

Gesture Mapping: Certain facial expressions are translated into commands or cursor
movements. For instance, a smile could be connected to a mouse click, while raising an
eyebrow could be connected to an upward cursor movement.

Real-Time Tracking: To make sure the cursor control is responsive and in line with the user's
gestures, the system continuously monitors and analyzes the user's face motions in real time.

Calibration and Personalization: Depending on the user's unique facial characteristics,

expressions, and comfort level, the system may have a calibration phase when users can
customize the gesture mapping and change parameters.

User Interface: The interface is designed to be simple and easy to use, with visual cues and
feedback to help users grasp and manipulate the cursor movements with their facial gestures.
Through the provision of an accessible and alternate input method, the proposed system
seeks to empower people with physical disabilities or limits to engage with computers and
carry out tasks with greater independence. The method provides a natural and easy approach
to operate the cursor by utilizing facial motions, which could improve user experience and
encourage inclusion in computer interactions.

Requirements for Software and Hardware

Hardware Requirements
Processor - Intel® Core™ i3 or higher
Speed - 1.3Ghz or higher
RAM - 4 GB or higher
Hard Disk - Minimum 32 GB
Software Requirements
Windows 7 or higher, LINUX OS
Computer vision library - OpenCV
Camera – Webcam
Programming Language – Python

Algorithm
The Haar Cascade algorithm is a machine learning-based object detection method
used for identifying objects in images or videos. It's particularly popular for face detection
but can be trained to detect other objects as well. Haar-like features are rectangular filters
applied to subsections of an image to capture specific patterns. These features include edge
features, line features, and more complex pattern. Images that are positive and negative are
used to train the algorithm. While negative photos lack the object to be recognized, positive
images do. The algorithm learns from these images to identify patterns that represent the
object of interest. The algorithm employs a cascade of classifiers, where each stage filters out
negative regions of an image that are unlikely to contain the object. This helps in faster
processing as fewer regions are scrutinized in subsequent stages. The algorithm moves a
window of varying sizes across the image, applying the learned features to each window to
determine if it contains the object being detected. Integral images are used to rapidly compute

MZUJHSS, Vol. X, Issue 1, June 2024 271

the Haar-like features, which speeds up the detection process. Haar Cascades can be used for
facial detection, allowing the system to identify and track facial features such as the eyes,
nose, and mouth.

Here’s a step-by-step explanation of the Haar cascade algorithm for face detection.

Data Collection:
Positive Images: Collect a large number of images containing faces. These are images where
the object of interest (faces) is present.

Negative Images: Gather images without faces or with different backgrounds. These are used
to train the classifier to distinguish between faces and non-faces.

Feature Extraction: Haar-like features are used to represent facial characteristics. These
features are rectangular patterns that capture local intensity changes in the image. Examples
of Haar-like features include edge features, line features, and center-surround features. Each
feature represents the difference in intensity between adjacent rectangular regions of the
image.

Training the Cascade Classifier:

Initialize the cascade classifier: Train the classifier using a machine learning algorithm, often
AdaBoost (Adaptive Boosting). Ada Boost selects a subset of Haar-like features that are most
effective at distinguishing between positive and negative examples. At each iteration,
AdaBoost assigns weights to the training samples based on their classification error.
Misclassified samples are given higher weights to focus on them in subsequent iterations. The
process continues until a predefined number of iterations or until a desired level of accuracy
is achieved.

Creating the Cascade: After training, the selected features are organized into a cascade of
classifiers. Each stage of the cascade consists of a subset of features. During detection, the
image is scanned using a sliding window technique, and at each window position, features are
evaluated. If a window passes all stages of the cascade, it is classified as a detection.
Otherwise, it is rejected, saving computational resources.

Adaptive Thresholding: Apply adaptive thresholding to handle variations in lighting

conditions and backgrounds. Adjust the threshold dynamically based on the local region of
the image, ensuring robustness to changes in illumination.

Face Detection: Apply the trained cascade classifier to new images. Slide the detection
window across the image at different scales and positions. At each window position, evaluate
the presence of facial features using the cascade of classifiers. If the features match closely
with facial characteristics, classify it as a detection. Accumulate multiple detections to
identify potential faces in the image.

MZUJHSS, Vol. X, Issue 1, June 2024 272

Post-processing: Refine the detected faces using techniques such as non-maximum
suppression to eliminate overlapping detections. Optionally, apply additional algorithms for
facial landmark detection or face recognition.

Integration and Deployment: Integrate the face detection module into the desired
application, such as security systems, video surveillance, or facial recognition software. Test
the performance of the system on different datasets to evaluate its accuracy and efficiency.
Deploy the system for real-world applications, ensuring scalability and reliability.

Modules
Face Detection
Facial landmark detection typically follows a two-step process. The first step involves
pinpointing the position of the face in the given image. This is achieved through various face
detection algorithms, which utilize computational models to locate and outline the facial
region. Once the face is identified, the second step involves detecting and localizing the key
facial structures within the ROI. This includes identifying the precise locations of eyebrows,
eyes, mouth ends, nose, and jawline

Fig. 2: Face Detection

Eye Blinking Detection

The process of detecting eye blink using facial landmarks begins by identifying the
left corner of the eye, followed by a clockwise exploration of the surrounding area. Each eye
is represented by coordinates (x, y) that indicate its position. When analyzing an image, it is
crucial to focus on specific key points of the eye. Researchers Soukupova and Cech
introduced a method in their paper titled "Real-Time Detection of Eye Blink Using Facial
Landmarks." They derived an equation known as the eye aspect ratio (EAR) that captures the
relationship between the height and width of these key points (coordinates).

Fig. 3: Eye Region Detection

MZUJHSS, Vol. X, Issue 1, June 2024 273

In the context of eye region detection using Haar cascades p1, p2, p3, p4 and p5 refer to the
stages of the cascade classifier. These stages are integral parts of the cascade classifier used
in the Viola-Jones face detection framework.

P1: This represents the first stage of the cascade classifier. In this stage, a subset of Haar-like
features is applied to the region of interest (ROI) in the image. These features are simple
rectangular patterns that capture variations in pixel intensity. The purpose of this stage is to
quickly discard regions of the image that are unlikely to contain the object of interest (in this
case, eyes), thereby reducing the computational load for subsequent stages.

P2: If a region passes the first stage, it moves on to the second stage of the cascade classifier
represented by P2. In this stage, another subset of Haar-like features is applied to the region.
These features are more complex and specific to the characteristics of eye regions. The
classifier further refines its decision on whether the region contains eyes or not based on the
responses of these features.

P3: Regions that pass the second stage proceed to the third stage represented by P3 Here, the
classifier applies additional Haar-like features that are even more discriminative for detecting
eyes. The purpose of this stage is to achieve a higher level of accuracy in distinguishing
between eye regions and non-eye regions.

P4: If a region successfully passes through the third stage, it undergoes evaluation in the
fourth stage represented by P4. This stage typically consists of a more comprehensive set of
Haar-like features specifically designed to capture subtle details and variations in eye regions.
The classifier makes a more refined decision based on the combined responses of these
features.

P5: The fifth stage P5 represents the final stage of the cascade classifier for eye detection.
Regions that pass through all previous stages are subjected to a final set of Haar-like features
tailored to the specific characteristics of eye regions. The classifier makes the ultimate
decision on whether the region contains eyes based on the collective responses of the features
in this stage.

Each stage of the cascade classifier acts as a filter, progressively narrowing down the search
space and reducing false positives while maintaining high detection rates. By using multiple
stages, the classifier can efficiently detect eyes in images with varying backgrounds, lighting
conditions, and occlusions.

Mouse Event Control Mechanisms

One approach to address this is by utilizing assistive technologies such as eye tracking
systems. These systems track the movements and positions of the user's eyes, allowing them
to control mouse events through eye movements and blinks. By precisely detecting and
interpreting eye movements, the cursor on the screen can be manipulated, replicating the

MZUJHSS, Vol. X, Issue 1, June 2024 274

functionality of a traditional mouse. Eye tracking technology, combined with facial landmark
analysis, plays a crucial role in achieving accurate control over mouse events for amputees.
By utilizing facial landmarks to track the position and movement of the eyes, the system can
translate these movements into cursor movements on the screen.

In addition to cursor movement, clicking actions can also be facilitated through eye blink
detection. By monitoring intentional blinks, the system can be programmed to interpret
specific blink patterns as mouse clicks. This allows amputees to interact with software
applications, browse the internet, click on icons, and perform various tasks that typically
require mouse clicks. The implementation of these control mechanisms requires robust
algorithms and software that can accurately detect and interpret eye movements and blinks.
Machine learning techniques may be employed to train models that can adapt to individual
users' eye characteristics and optimize the accuracy of cursor control and clicking actions

Fig. 4: Mouse Event Control

Image Processing
Controlling mouse events for amputees involves the utilization of image processing
techniques to analyze facial features and extract relevant information for gesture recognition
and control. Image processing plays a vital role in capturing and interpreting facial
expressions and gestures. By analyzing facial features, such as the position of the eyes,
eyebrows, mouth, and other key landmarks, it becomes possible to identify specific gestures
and translate them into corresponding mouse events. To enable gesture recognition and
control, a combination of computer vision algorithms and machine learning techniques can be
employed. These algorithms process the input from a camera or other imaging devices to
detect and track facial features in real-time. The extracted information is then analyzed using
machine learning models to recognize predefined gestures or patterns. For example, a raised
eyebrow or a wink could be recognized as specific commands, such as a mouse click or a
right-click. By mapping these gestures to corresponding mouse events, amputees can interact
with digital interfaces without the need for physical mouse input. In addition to gesture
recognition, image processing can aid in the customization and personalization of the control
system. Each amputee may have unique facial features or specific requirements, and image
processing algorithms can adapt to these individual characteristics.

MZUJHSS, Vol. X, Issue 1, June 2024 275

 Fig. 5: Image Processing

Experimental Work
Experimental Setup
Data Collection: Gather a dataset consisting of facial images and videos capturing diverse
head movements, eye gestures, and facial expressions. Include data from individuals with
different physical abilities to ensure the system's adaptability.

Pre-processing: Pre-process the collected data by annotating facial landmarks (eyes, nose,
mouth) and segmenting facial regions for training and testing purposes.

Model Training
Facial Landmark Detection: Train a facial landmark detection model using machine learning
techniques (e.g., CNNs with libraries like dlib or OpenCV) to accurately identify key facial
points in real-time from webcam feeds.

Head Movement Interpretation: Develop algorithms to interpret head movements by

analyzing the changes in facial landmarks over time. Implement logic to convert head
gestures into corresponding mouse cursor movements.

Eye Tracking: Train a model for eye tracking, particularly iris detection, to monitor and
interpret eye movements. This information will aid in controlling specific mouse actions
based on gaze direction.
System Integration
Interface Development: Create an interface that displays the webcam feed in real-time,
overlays visualizations of detected facial landmarks, and demonstrates cursor movements
based on interpreted head and eye gestures.

Real-time Processing: Implement the system to process facial movements in real-time,

updating cursor movements instantly to ensure a responsive user experience.

MZUJHSS, Vol. X, Issue 1, June 2024 276

Evaluation
Performance Metrics: Define performance metrics such as accuracy in head movement
interpretation, precision in eye tracking, and responsiveness of the cursor control system.

Testing Scenarios: Conduct testing sessions with individuals with physical challenges to
evaluate the system's usability, reliability, and accuracy in controlling the mouse cursor
solely through facial movements.

Feedback and Iteration: To find areas for enhancement and more algorithmic refinement,
collect user input.

Output Screens

Fig. 6: Cursor Movement Fig. 7: Left Click

Fig. 8: Right Click Fig. 9: Scroll Mode

Integration and Experimental Results

This typically includes integrating the facial gesture recognition algorithm, cursor
control logic, and user interface elements. The integration process ensures that these
components work together seamlessly to provide accurate and intuitive cursor control based
on facial gestures. This experimental study evaluates the effectiveness and usability of a
head-controlled mouse system for human-computer interaction. The system allows users to
control the mouse cursor using head movements, offering an alternative input method for
individuals with limited mobility. The study aims to assess the performance, user
experience, and potential applications of this technology.

MZUJHSS, Vol. X, Issue 1, June 2024 277

Methodology
Participants: The study recruited 20 participants with varying levels of computer proficiency
and no prior experience with head-controlled mouse systems.

Task Design: Participants performed a series of common computer tasks, including

navigating a web page, clicking on specific elements, and typing text.

Evaluation Metrics: The study measured task completion time, accuracy, user satisfaction
(via surveys), and subjective feedback on usability and comfort.

Results
Performance: Participants demonstrated a learning curve in using the head-controlled mouse,
with improved accuracy and speed over time.

Task Completion Time: On average, participants took slightly longer to complete tasks
compared to traditional mouse input but showed comparable performance for simple tasks.

Accuracy: Participants achieved high accuracy rates in selecting targets with the head-
controlled mouse, indicating its precision.

User Satisfaction: Overall, participants reported positive experiences with the head-controlled
mouse, highlighting its potential for improving accessibility and independence in computer
use.

Usability and Comfort: Participants provided feedback on the comfort of wearing the head-
tracking device and suggested improvements in ergonomics and calibration.

SN Modules If processed If not

processed
1 Face Detection User’s face is successfully User’s is not detected due to
detected some external difficulties like
bad lightning.
2 Eye Blinking User’s various face gestures User’s gestures are not
Detection are recognized. recognized.
3 Image Processing This dataset may include Left and right clicks are not
images of individuals with performed.
different amputation levels of
performing various facial
gestures to control mouse
events.
4 Mouse Event Scroll mode is activated when Scroll mode is not Activated
Control the user opens the eyes and
Mechanisms mouth for few seconds.

MZUJHSS, Vol. X, Issue 1, June 2024 278

Discussion
The experimental results suggest that head-controlled mouse systems have the
potential to enhance human-computer interaction for individuals with motor impairments.
While participants adapted to the technology and achieved satisfactory performance, further
research is needed to optimize usability, reduce setup time, and address ergonomic concerns.

Testing
Functional Testing: Verify that the system accurately recognizes and interprets facial
gestures, performs cursor movements, clicks, scrolling actions, and executes predefined
commands as intended.

Conclusion
By using head and eye motions to control the mouse pointer, the suggested method
seeks to provide physically challenged people with an alternate mode of computer access.
This system emphasizes the utilization of face gestures for enhanced accessibility, mobility,
and reliability in user-computer interaction by utilizing the webcam without the need for
extra hardware. This method acknowledges head movement as an important real-time input
medium, particularly important for people with physical limitations. With the integration of
iris tracking and head position-corresponding cursor movement, the system presents a
revolutionary approach to eye control. Constructed with Python and machine learning
methods, the system uses the user's face motions alone to offer a straightforward and practical
interactive mode. To sum up, the technology is innovative because it uses facial gestures that
are recorded by a webcam to control the cursor on a computer screen, giving physically
challenged persons a dependable and accessible way to operate the cursor.

Future Scope
Here are some potential future directions for research or development in the field of
human-computer interaction (HCI) based on head-controlled mouse projects:

Enhanced Gesture Recognition: Improve the accuracy and robustness of gesture recognition
algorithms to allow for more precise and intuitive control of the head-controlled mouse. This
could involve exploring advanced machine learning techniques, incorporating user feedback
mechanisms for real-time adaptation, or integrating multiple sensors for better gesture
detection.

Adaptive User Interfaces: Develop adaptive user interfaces that dynamically adjust their
layout, functionality, or control mechanisms based on the user's interaction patterns,
preferences, and abilities. For instance, the interface could automatically optimize cursor
speed or sensitivity based on the user's head movement characteristics.

Multi-Modal Interaction: Explore the integration of head-controlled mouse systems with

other modalities such as voice commands, eye tracking, or brain-computer interfaces (BCIs)
to create more versatile and natural interaction experiences. This could enable users to
perform complex tasks more efficiently by combining different input modalities.

MZUJHSS, Vol. X, Issue 1, June 2024 279

Accessibility and Inclusivity: Focus on designing HCI systems that are accessible to users
with diverse abilities and disabilities. This could involve conducting user studies with
individuals with specific needs (e.g., motor impairments) to identify usability challenges and
develop tailored solutions to improve their interaction experience.

Virtual and Augmented Reality Integration: Investigate the integration of head-controlled

mouse systems with virtual reality (VR) and augmented reality (AR) environments to enable
hands-free interaction in immersive computing environments. This could open up new
possibilities for applications such as virtual training simulations, immersive data
visualization, or collaborative virtual environments.

Real-Time Feedback and Assistance: Develop intelligent systems that provide real-time
feedback and assistance to users during interaction with the head-controlled mouse interface.
This could include proactive error correction, context-sensitive help prompts, or adaptive
assistance based on the user's task and performance.

Long-Term User Adaptation: Explore techniques for long-term adaptation and

personalization of head-controlled mouse interfaces based on the user's evolving preferences,
habits, and motor abilities. This could involve continuous learning algorithms that adapt the
system's parameters over time to optimize user comfort and efficiency.

Ethical and Social Implications: Investigate the ethical and social implications of head-
controlled mouse technologies, including issues related to privacy, autonomy, and the impact
on social interactions. This could involve interdisciplinary research combining HCI with
fields such as ethics, psychology, and sociology to ensure that the technology is developed
and deployed responsibly.

*******

References
[1] Wahyu Rahmaniar, Alfian Ma’arif, “Touchless Head-Control (THC): Head Gesture
Recognition for Cursor and Orientation Control “The 11th International Conference on
Computing, Communication, and Networking Technologies which took place in 2020.
(ICCCNT).

[2] Atsunori Moteki, “Head gesture 3D interface using a head mounted camera”, 2014 IEEE
International Symposium on Circuits and Systems (ISCAS).

[3] Sarah Chaudhry, Muhammad Usman Ghani, "Gaze Pointer: A Real Time Mouse Pointer
Control Implementation Based on Eye Gaze Tracking”
([Link] [Link]). (February
6, 2014).

MZUJHSS, Vol. X, Issue 1, June 2024 280

[4] Nieman, Schmidt, Jochsen, Vogt, and "A structure-from-motion approach for calibration-
free hand-eye calibration." Springer Berlin Heidelberg, 2005. Pattern Recognition, 67–74.)
(2005) (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 9, No. 9, 2018.

[5] Syed Muhammad Tahir Saleem, Sammat Fareed, Farzana Bibi, Arsalan Khan, Shahzad
Gohar, “IMouse: Eyes Gesture Control System”, (IJACSA) International Journal of
Advanced Computer Science and Applications, Vol. 9, No. 9, 2018.

[6] Carlos H. Morimoto, Marcio R.M. Mimica, “Learning Tennis through Video-based
Reflective Learning by Using Motion Tracking Sensors Computer Vision and Image
Understanding”, Volume 98, Issue 1, April 2005, Pages 4-24.

[7] Carlos H. Morimoto, “Detecting Eye Position and Gaze from a Single Camera and Light
Sources”, Publication in research gate.

[8] Bozomitu, Alexandru Păsărică, Vlad Cehan, Cristian Rotariu, “Pupil Centre Coordinates
Detection Using the Circular Hough Transform Technique”, Publication in research gate.

[9] S. Sharanya, Yazhini.K, Madhumitha.R. P, Yamuna Rani, “Eyeball Cursor Movement

Detection Using Deep Learning”. Publication in research gate.

[10] Chih-Hung Yu1, Cheng-Chih Wu1, Jye-Shyan Wang, Hou-Yu Chen and Yu-Tzu Lin,
“Learning Tennis through Video-based Reflective Learning by Using Motion Tracking
Sensors”. Publication in research gate.

MZUJHSS, Vol. X, Issue 1, June 2024 281