A new eye gaze detection algorithm using PCA features and recurrent neural networks

Abstract

The paper presents a new eye-gaze detection algorithm from low resolution images using Principal Component Analysis (PCA) and recurrent neural networks (RNN). First, eye images are extracted from human face images using Adaboost classifier and Haar-like features. A set of sample eye images captured under different lighting conditions is used to build an eigeneye space based on PCA. The coordinates of the sampled eye images in the eigeneye space are employed to train three-layer recurrent neural networks. Experimental results show that the trained neural networks can determine eye gaze direction with high accuracy and robustness to lighting conditions of the working environment.