Sign language recognition with computer vision

Tee, Wei Heng (2024) Sign language recognition with computer vision. Final Year Project, UTAR.

Preview

PDF Download (4Mb) | Preview

Abstract

This project centres on crafting a web application that recognizes sign language, accessible across different devices and operating systems. Its primary objective is to convert American Sign Language (ASL) gestures into text, facilitating communication for the hearing impaired. Leveraging Computer Vision (CV), the project aims to equip computers with the ability to interpret visual cues. Addressing the inherent challenges of sign language, including its limited universality and usage, the project unfolds in seven key stages: data collection, feature engineering, data preparation, model design and training, testing and evaluation, and application development. Noteworthy testing results showcase a robust 95% training accuracy and an 85% testing accuracy. The envisioned web application boasts a feature-rich interface encompassing gesture recognition, user ratings, translation history management, account administration, and an extensive sign language dictionary. Clarifying the system's functionality, diagrams are employed for enhanced comprehension. In the implementation phase, meticulous attention is paid to hardware and software configurations, with detailed setup instructions provided. System operations are thoroughly elucidated, alongside candid discussions on encountered challenges such as data quality issues and processing constraints. To ensure the system's reliability, a comprehensive testing regimen is executed, spanning video frame capture, model prediction, and web application feature validation. In summary, this project marks a good stride towards enhancing communication accessibility for the hearing impaired.

Actions (login required)