A computer vision-based system for recognition and classification of Urdu sign language dataset

Area of Science:

• Computer Science
• Artificial Intelligence
• Linguistics

Background:

• Social communication is vital, with sign language bridging the gap between deaf and non-deaf communities.
• Approximately 70 million deaf individuals globally use around 300 distinct sign languages.
• Urdu Sign Language (USL) is a visual communication system for daily interaction.

Purpose of the Study:

• To create a dataset of Urdu Sign Language (USL) images.
• To develop and test a machine learning classifier for automated USL recognition.
• To evaluate the performance of different machine learning models for USL number classification.

Main Methods:

• A dataset of 1,560 USL images was collected and photographed.
• A bag-of-words (BoW) paradigm was employed for automated identification.
• Support Vector Machine (SVM), Random Forest, and K-nearest neighbor (K-NN) classifiers were utilized with BoW histogram features.

Main Results:

• The Random Forest classifier achieved 88% accuracy.
• The Support Vector Machine (SVM) classifier attained 90% accuracy.
• The K-nearest neighbor (K-NN) classifier reached 84% accuracy.

Conclusions:

• The proposed machine learning strategy effectively automates Urdu Sign Language number recognition.
• Support Vector Machine demonstrated the highest accuracy among the tested classifiers.
• This system offers a promising solution for enhancing communication accessibility for the deaf community.