TY  - JOUR
AU  - T Layaraja
AU  - T Akhila
AU  - V Harshini Reddy
AU  - Y Raghunandhu
AU  - V Pranav Reddy
PY  - 2026
DA  - 2026/02/09
TI  - A Deep Learning-Driven Real-Time Eye Gesture Recognition Framework for Intelligent Hands- Free Multimedia Control Systems
JO  - Global Journal of Engineering Innovations and Interdisciplinary Research
VL  - 6
IS  - 2
AB  - Hands-free human-computer interaction is essential for accessibility, especially for individuals with
motor impairments, and enhances user experience in multimedia applications. Traditional input devices
like keyboards and mice limit usability in scenarios requiring contactless control. This paper proposes a
deep learning-driven real-time eye gesture recognition framework for intelligent hands-free multimedia
control systems. The system utilizes webcam-captured video streams to detect and classify eye gestures
(e.g., blink, double blink, gaze left/right/up/down) using a hybrid Convolutional Neural Network (CNN)
and Long Short-Term Memory (LSTM) architecture, augmented with MediaPipe for facial landmark
extraction and Eye Aspect Ratio (EAR) computation. Recognized gestures map to multimedia commands
such as play/pause, volume adjustment, next/previous track, and fullscreen toggle. The framework
ensures low-latency inference (<50 ms per frame) on standard hardware. Experimental evaluation on
custom and public datasets (e.g., EYEDIAP, custom multimedia control gestures) demonstrates high
accuracy (96.5%), precision (95.8%), recall (96.2%), F1-score (96.0%), and robustness to lighting
variations and head pose changes. The proposed system promotes inclusive, intuitive control for media
players, smart TVs, and virtual environments while maintaining privacy through on-device processing.
SN  - 3066-1226
UR  - https://dx.doi.org/10.33425/3066-1226.1212
DO  - 10.33425/3066-1226.1212