REAL-TIME HAND TRACKING USING THE VIOLA AND JONES METHOD

摘要

In this paper a method to generate examples for training a classifier based on the Viola and Jones algorithm is introduced. By using this method, positive examples are generated to train classifiers able to distinguish a certain angle of rotation that is normal to the image plane. The examples were used to train and assess classifiers that are able to track and recognise hands posed in a single gesture. Experiments compared the performance of classifiers trained at different angles. The results show that the Viola and Jones algorithm has the potential to be used for more sophisticated gesture and hand recognition, but it also points out to the difficulties of using the method generically for rotated/articulated objects.