SignConnect:

Sign
Language
Web
Conferencin
g Tool
SignConnect:
Sign
Language
Web
Conferencin
g Tool
 1
Background 6 Methodology

Problem Tools &

Objectives 3
Table of 8 Design
Content
Significanc
4 9Timeline
e
Literature
Review
5 Conclusion
10
Background

Global Community
DHH Community
Proble
m
Statem
ent
 Reduce
Inequality

Motivation

Good health
&Well Being
 Health

Education

Redue
Inequality
Objectives
 Objectives
General Specific
• Research and Development • Real-time Sign Language
• User Engagement and Feedback Translation
• Scalability and Sustainability • Customizable Avatar
Options
• Ethical Considerations
• Seamless Integration
with Video Platforms
• User-Friendly and
Accessible Design
• Support for Multiple
Significan
ce
 Related
Works
Overview
 Project Model Accuracy Advantage Disadvantage

American Sign Language Convolutional Neural 99% The system was able to limited number of ASL
Recognition Using Deep Networks (CNN) with the recognize isolated and gestures
Learning and Computer Inception continuous signs which made
Vision. it suitable for various use
Recurrent Neural Network cases.
(RNN) model known as Long
Short-Term Memory (LSTM)

Deep Learning-Based Histogram of Oriented 94.3% the CNN model achieved the requires a large amount of
Approach for Sign Language Gradients (HOG) and highest accuracy for labeled data for training the
Gesture Recognition with Convolutional Neural recognizing ASL alphabets. deep learning models, which
Efficient Hand Gesture Networks (CNN). can be time-consuming and
Representation. expensive.

Indian Sign Language multi-layered CNN Indian Sign Language - a multi-layered CNN-based limited research that was
Gesture Recognition Using architecture for static 99.08% architecture proved to carried out in recognition of
Image Processing and Deep gestures and LSTMs with Detection of American Sign provide excellent results and ISL gestures
Learning. convolutional kernels for Language - 97.71% set new benchmarks.
dynamic gestures.

Real-Time American Sign Hidden Markov Models 94.4% Combination of HMMs with Has a limited vocabulary of
Language Recognition from (HMMs) feature extraction only 20 ASL signs, which
Video Using Hidden Markov techniques helps in handling may not be sufficient for
Models.” variations due to noise, practical applications that
lighting, and individual require recognition of a
differences in signing styles. larger vocabulary.

Hand Gesture Recognition No specific model mentioned No specific accuracy Deep Learning technique has Deep learning techniques
based on Computer Vision mentioned however, a high the ability to learn complex require a large amount of
accuracy was achieved with features datasets
deep learning technique

Sign Language interpreter The proposed model is a 95% recognition rate for If a deaf person If the user does not perform
system system: An glove equipped with five flex static gestures communicates using the SL the gestures accurately, the
alternative system for sensors, interfacing with an that he is familiar with, then system may not be able to
Methodolo
gy
Block
Diagra
m
Mod
els Gesture Posture
Landmark
Recognition Recognition
Detection 3D CNN + MediaPi
MediaPi Transformer based pe
pe models
● Contains hand ● 3D CNN for capturing ● Contains a model
model called the temporal called MediaPipe
Mediapipe hands information of the Pose which can
that can be used to gesture and will be fed be used for
extrand hand into a Transformer posture
model recognition
UI/UX
React

Pytho
n

MongoDB

Figma

GitHub
10. **Conclusions:**
- Summarize the main points discussed in the
presentation, highlighting the significance of Sign Connect
and its potential impact.
- Reiterate the importance of improving communication
accessibility for the deaf and hard of hearing community.
- Invite questions and feedback from the audience to
encourage engagement.
Conclusion