SmartAlign: A Posture Correction

Application
Submitted by

NAME SECTION
M. Umair Amjad FA-21/BSCS/404-G
Shahwaiz FA-21/BSCS/405-G

Supervised by

Miss Ayza Batool

Department of Computer Science

Lahore Garrison University, Lahore
Abstract
In this paper, the Software Requirements Specification (SRS) for a mobile application called
"SmartAlign" is presented. This application aims to break new ground in posture correction.
Integrated into it, an AI computer vision and machine learning that will analyze the users' posture
in real-time has corrective feedback mechanisms and other personalized exercises. The main aim
is to make it attractive and accessible to all age brackets interested in improving posture and the
general well-being.

Introduction

Purpose
SmartAlign is a posture correction application that aspires to intelligent posture correction
through advanced AI technologies in real time and for constant monitoring and improvement in
user posture.These will reflect imperfections associated with body posture and these will guide
the user through corrective actions via visual or audible clues.

Target Group
The target audience for this document includes the following stakeholders:

 Development team - to guide implementation and integration of all artificial intelligence

technology in the software design.

 Project Managers: Monitoring the project on its progress while ensuring all requirements
are satisfied.

 Testers: Testing for the functionality and usability of the system.

 Investors and Stakeholders: Giving a view of the project, its objectives, functionalities,
and market opportunities.

Product Scope
SmartAlign shall be distinguished from traditional posture correction applications
in the following ways:

 Real Time Posture Guidance: Immediate feedback delivered onto the user posture
example as improved suggestion.

 AI-Based Analysis: State of the art machine learning methods in analyzing posture.

 Custom Program Recommendations: Building up personalized exercises and

improvement plan based on individual needs.

 User Engagement Features: Gamification and progress tracking will form an integral
part in promoting frequent usage of the application.

Problem Definition

Problem Statement

Excessive sitting, excessive time spent in front of a screen, and lack of exercise tend to be
common situations these days, and all these prerequisites worsen back pain, neck strain, etc., and
misalignment of the spine.Many people are not conscious of their bad postures, making self-
assessment and self-correction almost impossible.

Key Challenges
 Awareness: The user may not recognize that he/she is exhibiting poor postures.

 Self-Assessment: Assessing one's own postures can be a difficult task.

 Consistency: Even with input, users find themselves unable to maintain good postures.
  Personalization: Generic solutions do not fit a person's own specific needs with respect to
a particular posture.

The Offered Solution

SmartAlign helps by:

 Real-time AI-enabled Posture Monitoring: Continuous observation and

instantaneous feedback.

 Custom Recommendations: Personalized exercises for a posture correction.

 Awareness: Awareness cannot be achieved by general means, for it is not an easy task to
achieve some uncommon posture in isolation.

The Answer That Is Proposed

SmartAlign handles the problems with:

 Real-time, AI-enabled Posture Monitoring, that is a continuous observation followed up

by feedback instantaneously.

 Individual Suggestions, that is, of specific exercises for the correction of posture.

 Awareness: No one can ever really achieve an awareness since such things can be
personal and do not come to everyone to achieve some uncommon strange posture alone.
The Solution Offered

SmartAlign helps in:

 Real-time AI-enabled Posture Monitoring: Continuous observation and

instantaneous feedback.

 Custom Recommendations: Personal exercises in posture correction.

 Now that one realizes awareness: no one is going to achieve awareness since those
things could be personal and do not come to everyone to achieve some uncommon strange
posture themselves.

The Answer That Is Proposed

SmartAlign handles the problems with:

 Real-time, AI-enabled Posture Monitoring, that is a continuous observation followed up

by feedback instantaneously.

 Individual Suggestions, that is, of specific exercises for the correction of posture.

 Engaging User Experience: Gamification elements to promote habit formation.

 Educational Resources: Information on proper posture habits and benefits.

Software Requirement Specification

User Interfaces
 Login/Sign-up Screen
 .

Home Screen:
  Exercise Screen:

Software Interfaces
 Computer Vision Green Libraries (OpenCV, MediaPipe), pose or other image processing
features.
 Machine Logical Models PoseNet , EfficientPose To specify the postural changes.
 Data Storage local and cloud user profiles and tracking history.

Communication Interfaces
 Camera access captures the posture of the user.
 Network connectivity gives one access to cloud processing and updating.

System Features
 Analysis by AI posture: Poor posture is detected and feedback given on it.
 Exercise Programs customize for the individual.
 Tracking of Progress can visualize improvement over time.

Nonfunctional Requirements Specifications.

 Minimum Latency Performance for Real-time Feedback.
 Security the Data of the User is Safe.
 Usability a Friendly Design Accessible for All Ages.

Methodology
 SmartAlign will utilize an iterative development process.
 Prototype Development-MVP Building Competition Test.
 AI Model Training Data Collection for Model Training Precisely on the Target.
 User Testing and Feedback Using User Input to Refine the Feature.
 Detailed Design and Architecture Layer of AI Processing
 Pose Estimation Module Detects by a Computer Vision System Key Joints.
 Posture Classification Module Identifies Correct and Incorrect Posture.

Data Flow
1. Capture by cam of an individual's posture.
2. AI processes the image and classifies it as living.
3. Corrective information is given to the user.
Implementation and Testing
Test Methodology includes
Unit testing for the verification of individual components as well as
Integration testing that ensures that the functionality operates smoothly across modules
User testing is the collection of feedback from prospective users

Evaluation metrics
 to evaluate the Impact Assessment of the
 Classification Accuracy evaluation of the AI model
 Latency at the system, measuring speed at which real time feedback is provided
 User Satisfaction that measures user comments in tests conducted to evaluate application
usability

Results and Discussion

Performance Findings
 Acts with high accuracy in direct provisions during real time
 Very minimal latencies recorded when feedback is given

Usability Results
 Pictures clearly describe posture in a user-friendly interface.
 System Reliability Stable function when used continuously.

Conclusion and Future Work

SmartAlign understands a large part of the problem's solution, which is posture, through AI
solutions. Future development will include:
 More additional training on the AI model to improve accuracy
 A voice command that will allow hands-free use
 Integrate AR features to provide immersive feedback in performance
Diagrams
1. System Architecture Diagram
Elaboration of the Posture Detection and Classification
Process
The Smart Align application is designed to monitor and correct body posture using computer
vision and AI-based techniques. It follows a structured workflow that includes capturing,
processing, classifying, and responding to the user's posture in real-time.

1. Capturing the User’s Posture

The process begins when the user interacts with the system and provides a video or image as
input. The system utilizes a video camera to capture the body posture, ensuring that the necessary
data is recorded for analysis.
 The camera captures key body landmarks such as the head, shoulders, spine, and leg
positioning.
 The captured image or video is then forwarded to the preprocessing module for refinement.

2. Preprocessing the Input

Once the media is captured, it is processed to remove noise and enhance quality before being
analyzed by the AI model. This step ensures that only relevant posture-related features are
extracted, increasing the accuracy and efficiency of classification.
 The preprocessing module optimizes the input for AI-based evaluation.
 It ensures that the system focuses on body alignment rather than unnecessary background
elements.

3. AI posture classification
The improved data is then transferred to the YOLOV9 AI Model, a deeplearning program that
has been taught for posture recognition and classification. AI model:
Evaluates the found body posture against preordained ergonomics criteria.
classifies the posture into two groups:
 Good Posture; No alert is set off.
 Poor posture – an alarm goes off and remedial measures are advised.
4. System Response to Posture Classification
No additional actions are needed if the posture is rated as good. The system, though, promptly
takes correcting measures once poor posture is found:

A. Alert & Advices

Realtime responses including ideas on how to improve their positioning is given by the user. This
guarantees immediate attention and over time encourages habit correction.

B. Data storage & reporting

For future use, the system keeps posture data in a database. Each time a report is made, it is
stored in the database so users can monitor their progress over comparison.
 Users can get their posture analysis reports any time they want to.
 The system preserves a history of former postural evaluations to enable users track
consistent problems or progress.

C. feedback given in real time

The system gives realtime corrective feedback by:
 Showing on the screen of the user_git visual signs.
 Providing guidance in written form or videos showing the appropriate positioning.
2. Database Diagram
Smart Align Posture Detection System Database Structure

User Data Table

With core user data—name, age, gender, email, and credentials—stored, the User Table is the
central table in the system. With the promise of each user's data being secured, this table sets the
foundation for the interaction of users within the app.

Posture Evaluation Chart

Posting posturerelated data—images, video, say—enables the user to interact with the Posture
Analysis Table. Artificial intelligence models are used by the system to structure this data in
order to classify posture as good or bad. Shoulder assessments are retained here in the main
repository, which is essential for further exploration.

Recommendations table
The Recommendation Table gives remedies for users to improve their posture should bad posture
be identified. By means of the user's posture data, these AIgenerated suggestions are meant to
give custom feedback and hence improve the general training outcome.

Report Lists
The Report Table saves all the posture assessment so the users can write detailed reports. These
reports provide valuable data of their posture past that enables them to track evolution over time.
Users can rely on their earlier assessments at any time due to the data maintained here.

Development chart
Monitoring longterm posture change is the responsibility of the Progress Table. The system is
capable of giving AIdriven feedback through monitoring patterns in user data maintained under
continuous observation. Through continuous evaluation and progress tracking, this table plays a
key role in helping users maintain more upright position habits.
3. Use Case Diagram

User Authentication:

Posture Analysis:

Report & Feedback:

4. Sequence Diagram

 User Authentication
A sequence diagram for user authentication
The sequence diagram is the process for authenticating the user in the Smart Align posture
detection system. The three entities involved are User, App, and Database.

Process Overview

1. Application Opening:
 The process is started with the user's interaction by opening the app of the device.

2. Entering Credentials:
 The user provides login credentials, which generally consist of a password together with an
email, or registers providing input required such as name, age, gender, and password.

3. Credentials Verification:
 The application sends the user's credentials to the database for validation.

4. Database Response:
 The database checks the credentials provided against corresponding records stored in the
database and responds back to the app.
 If they are equal, then response = 'Valid'; Otherwise, response = 'Invalid'.

5. Success/Failure of Login:
 The application would notify the user regarding authentication failure or success depending
upon database response. Access to the system would be provided if successful; otherwise,
the user would be required to provide information again.

 Posture Analysis
 Posture Analysis Sequence Diagram
The sequence diagram gives the flow of posture analysis in Smart Align, showing four key
entities that are involved therein: User, App, AI Model, and Database.

Process Brief

1. Posture Selection:
 The user chooses the type of posture analysis to be performed- exercise posture, safe driving
posture pre-requisite, sitting posture or medical posture analysis.

2. Uploading of Image or Video:

 The user goes on to upload a picture or a video of his/her posture into the app.

3. Posture Processing using AI Model:

 The uploaded data is sent to the AI model (i.e. YOLOV9) by the app and is processed by the
model and categorized as good or bad posture.

4. Result Generation:
 The AI model returns the analysis result-good or bad posture to the app.
5. Storage of Results:
 The results of the analyses are stored in the database to be tracked in the future if the posture
is not good.

6. Display of Feedback and Suggestion:

 The app gives real-time feedback and suggestions for posture improvement based on the
analysis performed by the AI model.

 Report Generation
Generation of Report Sequence Diagram
The following sequence diagram shows the generation of a posture analysis report in the Smart
Align system. It consists of three major components: User, App, and Database.

Process Summary
1. User Request for Reports:
 The user requests to perform a posture analysis report through the application.

2. Retrieval of Analysis Data:

 The app fetches previously saved posture analysis results from the database.

3. Sending Retrieved Data:

 The database forwards the retrieved analysis data for the creation of a posture analysis
report.

4. Creating a Report to the Display:

 The app creates a report for users from retrieved data and shows it.

5. Saving Report for Post-Access:

 The created report is stored onto the database, thereby enabling the user to download or view
it another time when needed.
5. Data Flow Diagram (DFD)
Posture Analysis DFD
The DFD displays the process in which a user's posture is analyzed and reports or feedback is
given through AI-based classification. The flow originates from the user, and several steps are
present before the end output.

Process Flow
1. User Input
 The user uploads a video or photo of him/her/his/her postures for the system to analyze.

2. AI Processing
 The uploaded photo or video is examined by the YOLOv9 AI model to identify and examine
the posture of the user.

3. Posture Classification
 The detected postures are classified by the AI model as Good or Poor depending on specified
conditions.

4. System Response Depending on Classification

 If the posture is good, the user won't be alerted.
 If the posture is bad, a warning and actions to rectify the posture are given to the user.

5. Data Storage & Report Generation

 The report of posture analysis is saved in the database for the system record. The user can
download the report from the database for future reference.
6. ER Diagram

An Overview of Entity-Relationship Diagram (ERD)

The ERD depicts the database schema for posture analyzing. The entities covered in it are five in
number.
Entities and Attributes
1. User
 Stores user-related details like user_id, username, email, password, age, and sex.
 Primary Key (PK): user_id

2. Posture Analysis
 Involves details of the analysis, i.e., analysis_id, user_id, posture_type, media_type, file_url,
analysis_result, and feedback.
 Primary Key (PK): analysis_id
 Foreign Key (FK): user_id (links to User)

3. Recommendation
 Holds improvement suggestions, from which a recommendation is given for the posture
analysis.
 Primary Key (PK): recommendation_id
 Foreign Key (FK): analysis_id (links to Posture Analysis)

4. Report
 Reports generated for users based on analysis.
 Primary Key (PK): report_id
 Foreign Keys (FK): user_id, analysis_id (links to User and Posture Analysis)

5. Progress
 Keeps track of user progress in time by holding progress_id, user_id, analysis_id,
improvement_notes, and progress_percentage.
 Primary Key (PK): progress_id
 Foreign Keys (FKs): user_id, analysis_id (links to User and Posture Analysis)

Relationships
 User uploads the posture data into Posture Analysis.
 Posture Analysis gives recommendations for improvement through Recommendation.
 User creates posture reports stored in Report.
 User keeps track of improvements step by step in Progress.

Other UI Designs

Weekly/Daily Analysis

Weeky Insights
Clients Privacy