Published Paper Details:

In an era where computing is becoming om- nipresent, relying solely on physical peripherals like mice and keyboards presents real challenges, particularly re- garding hygiene in public spaces and accessibility for users with physical impairments. This paper introduces a contactless Human-Computer Interaction (HCI) system designed to overcome these barriers by merging computer vision with voice automation in a unified, efficient frame- work. Although several existing solutions rely on heavy deep learning classifiers such as CNNs or YOLO, our work demonstrates that a lightweight, geometric approach based on MediaPipe's hand landmarks can achieve comparable accuracy without the high computational cost. A key contribution of this research is the resolution of the "Blocking I/O" bottleneck--a common flaw in Python- based multimodal systems where voice recognition causes the video feed to freeze. We addressed this by engineering a concurrent, multi-threaded architecture that isolates the cursor control mechanism from the voice assistant. This ensures that the mouse cursor remains responsive even while the system is processing complex speech commands. Tests on standard hardware confirmed the system's via- bility, achieving a stable frame rate of over 30 FPS and a gesture recognition accuracy of 94%, making it a practical solution for sterile or accessibility-focused environments.