Human-computer interface using a web camera

[摘要] In this thesis we present a human-computer interface (HCI) system for disabledpersons using only a basic web camera. Mouse movements are simulatedby small movements of the head, while clicks are simulated by eyeblinks. In this study, a system capable of face tracking, eye detection (includingiris detection), blink detection and finally skin detection and facerecognition has been developed. A detection method based on Haar-like featuresare used to detect the face and eyes. Once the eyes have been detected,a support vector machines classifier is used to detect whether the eye is openor closed (for use in blink detection).Skin detection is done using K-means clustering, while Eigenfaces is usedfor face recognition.It is concluded that using a web camera as a human-computer interfacecan be a viable input method for the severely disabled.

[发布机构] Stellenbosch University