Real-Time Eye Blink and Facial Recognition System Using Computer Vision

Abstract

People who suffer from paralysis or severe motor disabilities often experience major difficulties when trying to communicate with others. Traditional communication tools such as keyboards, switches, and specialized eye-tracking devices are not always convenient because they may require physical movement or expensive hardware. This research introduces a real-time communication system that uses computer vision and deep learning techniques to interpret eye blinks and facial expressions. The system detects eye movements using dlib-based facial landmark detection and calculates the eye aspect ratio (ear) to determine blink events. Continuous analysis of ear values allows the system to identify intentional blinks and translate them into commands. In addition, facial expressions are recognized using a lightweight convolutional neural network (cnn) that processes 64×64 grayscale facial images and classifies them into seven emotional categories. The cnn architecture combines standard convolution layers with residual blocks and depth wise separable convolutions to reduce computational complexity while maintaining accuracy, along with batch normalization, relu activation, and global average pooling for efficient feature learning and classification. A virtual keyboard is incorporated to allow users to select characters using blink-based inputs. The recognized gestures are further used to generate voice feedback through a text-to-speech module. The proposed system works with a standard webcam and does not require specialized hardware devices. Experimental observations indicate that the system can provide an efficient and affordable communication interface for individuals with limited physical mobility.