DESIGN OF AN ASSISTIVE TECHNOLOGY FOR NIGERIAN SIGN ANGUAGE USING MACHINE LEARNING ALGORITHMS

The communication gap between Nigeria’s deaf community and hearing population motivated this project to create a machine learning-powered assistive system for Nigerian Sign Language users. The primary goal was to build a working prototype that could recognize sign gestures through camera, translate them into text, and produce spoken audio. The system focused on affordability and accessibility for educational settings where deaf students struggle to communicate with hearing teachers and classmates. A step-by-step waterfall development approach was adopted,building the system from scratch due to outdated existing solutions.Using a personal computer and standard webcam, the recognition pipeline was developed in python. Core components included OpenCV for capturing and processing video frames, a neural network model trained on sign gesture data, and Google’s text-to-speech service for audio output. Testing occurred in a local environment with a Streamlit web interface, though deployment challenges prevented wider distribution. The prototype correctly identified approximately 75% of tested sign gestures ,working best with clear, simple hand movement in good lighting. Complex signs with fast fingers motions proved harder to recognise. The system processed signs and produced audio within 1-2 seconds enabling natural conversation flow. While accuracy varied with lightning conditions and background clutter, the basic concept worked effectively .Based on these findings ,this project affirms that affordable sign language technology is achievable using everyday equipment and free software tools. Though limited by small training data and basic hardware, the prototype shows promise for helping deaf Nigerians communicates in classrooms and online meetings.