Sign to Urdu language Translator

Most people can listen and speak, and they use different languages to communicate. On the other hand, hearing impaired people use signs to communicate with other people. Urdu sign language is one of the preferable languages for deaf and dump people in whole world. To make the communication between d

Project Title

Sign to Urdu language Translator

Project Area of Specialization

Artificial Intelligence

Project Summary

Most people can listen and speak, and they use different languages to communicate. On the other hand, hearing impaired people use signs to communicate with other people. Urdu sign language is one of the preferable languages for deaf and dump people in whole world. To make the communication between deaf and normal people simpler and easier, Android-based Sign language to voice converter system are required. Therefore, the aim of our project is to make a bridge between Sign Understandable and non-sign understandable by developing an android based system which is cost effective and easy to use for everyone. The system has ability to convert different Urdu Sign language characters and convert it into speech. This system will be feasible in education sector for the fast learning of Urdu sign language to the normal people.Currently there is only one publicly available dataset for Urdu sign language therefore, we are using that dataset consisting of forty Urdu based characteristics. To overcome some computing issue initially we are working on ten Urdu Sign Language characteristics. The system recognizes almost all the given signs with good accuracy in real time. Urdu alphabet recognition dataset model has achieved 97.5% accuracy on test data and all label recognition this model also has achieved 98.2% accuracy on test data for applying Train Test Split over Dataset by 80% and 20%.For our system Different machine learning models were considered but we found Artificial Neural Network (ANN) to be the most suitable to accomplish these tasks.

Project Objectives

• No Android Based Urdu Sign Language Application Exist for deaf and dumb in real time environment. • Only one dataset available in JASON format for Urdu sign language character on Kaggle. • Android based application for deaf and dumb. • Make the sign language interpretation cheap by introducing an automated system that does not require any specialized sensors • Recognize the Urdu based characters in real time and convert it into voice. • Our software will recognize first 10 Urdu characters signs. • It is used to Train Normal People to teach deaf and dumb. 4 | P a g e • Reduced communication gap between normal and deaf people.

Project Implementation Method

Organize dataset for Urdu Sign Language: Firstly, we make a searching in lots of places for the sake of finding dataset of Urdu sign language , we find a dataset in Kaggle but unfortunately it is in .json format , and there is no description is mentioned that how to use the dataset .So , finally we decided to make our own dataset by applying preprocessing techniques.[6] 1. Classification 2. Clustering 3. Cleaning 4. Format data to make it consistent 5. Reduce data 6. Decompose data 7. Rescale data 8. Regression 9. Ranking Model Training Phase: • Train Model on dataset: By using CNN pretrained model we will train our dataset

Benefits of the Project

The aim of the project is to bridge the communication gap between signers and non-signers by developing an interpretation system which is cost effective and easy to use. The system is expected to have the ability to convert different Urdu signs into speech. The main objectives of this system are: • Make the sign language interpretation cheap by introducing an automated system that does not require any specialized sensors. • Provide an interactive learning system for Urdu Sign Language. • Make the system more accurate and easier to use.

Technical Details of Final Deliverable

Develop android app: We will make an android based application, along with integration of deep learning techniques and AI trained model. • Take input from Real time video capturing: We will use camera of mobile app that make a real time video and that it extracts the frames from video, then will apply background removing techniques and then identify the character from the frame and then generate an output. • Extract frames: We will extract the images in frames form by setting video in 60 frame/sec format then we extract 10th frame and compare it 9 passed frames if all 9 frames have same result then it would be provided to CNN model if any of passed 9 image would be different then we will extract that one passed frame. • Identify sign language character from frames: Provide frame will contain information in form of pixels. CNN models compare the features of that given frame if any match with provided trained model data then output will be showed in right format. • Generate output in the form of Urdu voice: We will generate some API’s of Android studio to generate voice of predefined label that is selected by AI model.

Final Deliverable of the Project

Software System

Core Industry

Education

Other Industries

Core Technology

Artificial Intelligence(AI)

Other Technologies

Sustainable Development Goals

Quality Education

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com