Hand Gesture Recognition

--
 --

Depth data acquired by current low-cost real-time depth cameras provide a more informative description of the hand pose that can be exploited for gesture recognition purposes. Following this rationale, we proposed a novel hand gesture recognition scheme based on depth information.

The basic framework [1,2] is the following. Color and depth data are firstly used together to extract the hand and divide it into palm and finger regions. Then different sets of feature descriptors are extracted accounting for different clues like the distances of the fingertips from the hand center, the curvature of the hand contour or the geometry of the palm region. Finally a multi-class SVM classifier is employed to recognize the performed gestures. Experimental results demonstrate the ability of the proposed scheme to achieve a very high accuracy on both standard datasets and on more complex ones acquired for experimental evaluation. The current implementation is also able to run in real-time.

We have also proposed several other solutions, including variations of the original approach, novel schemes and schemes based on different sensors:

• In [3] an improved version of the palm detection scheme is proposed.

• Different machine learning schemes are proposed in [4]  to improve the recognition accuracy

• Color-based descriptor are combined with the depth-based ones in [5]

• A review of several feature extraction schemes is presented in [6]

• The use of a Leap Motion sensor together with a depth sensor is discussed in [7,8]

• In a different approach [9] the curvature of the hand shape is used to separate the palm from the fingers. Then density-based clustering is used together with a linear programming approach to separate the various fingers

• The approach of [10], targeted to real-time ego vision systems, exploits descriptors based on the hand silhouette, in particular the curvature of the contour and features based on the distance transform.  Synthetic data obtained by an ad-hoc library made available on this website has been used for the training stage. The approach has been included into an augmented reality system including also an head mounted display presented in [11]

.

 

For all the publications: the copyright belongs to the corresponding publishers. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the publisher