Gesture labeling based on gaze direction recognition for human-machine interaction

Abstract

We propose a novel method of tracking the gaze direction of human eyes using a neural network (NN). First, ten primary parameters are extracted from the image of a human face by using the mountain algorithm and some other fast algorithms. With these primary parameters the feature parameters that are directly related to the gaze direction can be deduced. Then a NN is constructed to indicate the gaze direction after trained by the feature parameters. We have used this method to classify intentional hand gestures in a human-machine interaction system based on gesture recognition. The experimental results show that the proposed method is simple and convenient and can work effectively under changing light conditions.