Responsive human-computer interaction model based on recognition of facial landmarks using machine learning algorithms
Dhananjay Bisen,Rishabh Shukla,Narendra Rajpoot,Praphull Maurya,Atul Kr. Uttam,Siddhartha kr. Arjaria
DOI: https://doi.org/10.1007/s11042-022-12775-6
IF: 2.577
2022-03-08
Multimedia Tools and Applications
Abstract:This paper is addressed on the idea of building up a model to control computer systems by utilizing facial landmarks like eyes, nose and head gestures. The face recognition systems mainly detect and recognize eyes, nose and head gestures to control the movement of the mouse cursor in order to operate computer system in real time. This paper proposes the facial landmarks based human-computer interaction model in which histogram of oriented gradients (HOG) has been taken for global facial feature identification and extraction that is considered as HOG descriptors. Furthermore, pre-trained linear SVM classifier gets extracted features to detect whether it is a human face or not, including use of pyramid based images and sliding window algorithm. Moreover pre-trained ensemble of Regression Trees algorithm is applied to recognize facial landmarks such as eyes, eyebrows, nose, mouth, and jawline. The main purpose is to effectively utilize facial landmarks and allow the user to perform activities mapped to explicit eye blinks, nose and head motions using PC webcam. In this model, eye blinks has been detected through estimated value of eye aspect ratio (EAR) and newly proposed β parameter. Accordingly classification report has generated for both estimation and analysed best results for β parameter in terms of accuracy with 98.33%, precision with 100%, recall with 98.33% and F1 score with 99.16% under good lighting conditions.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering