Hand Gesture Detection and Recognition Using Affine-Shift, Bag-of-Features and Extreme Learning Machine Techniques

Abstract

This paper presents a real-time system for interaction with applications via hand gestures. Our system includes detection of bare hand from the video sequences by subtracting the background and including only the hand region. The system uses a machine learning approach. In the training stage the keypoints are extracted from the hand posture contour using the affine-scale invariance feature transform (ASIFT), and the keypoints are then clustered using the K-means clustering and mapped into the histogram vector (bag-of-features). Each vector is assigned a label which is treated as input to the Extreme Learning Machine (ELM) for training purpose. In the testing stage, for every frame captured using the webcam, the hand is detected, then, the keypoints are extracted from the hand segment only as described in our algorithm and fed into the cluster model to generate the vector and fed into the ELM training classifier to recognize the hand gesture. © Springer International Publishing Switzerland 2014.