Human Hand Gesture Recognition
 
 Abstract

We propose a novel 3D descriptor for hand gesture recognition for depth maps captured by Kinect-style cameras: Histogram of 3D Facets (H3DF). Different from previous methods, which massively used existing 2D image descriptors designed for 2D images on RGB channels or Grayscale channel, our descriptor starts from the substantial feature of a 3D object -- its surface. We first bring up with the idea of a 3D facet as a collection of discrete 3D cloud points. Then a normal-based coding process is performed to encode each facet into a compact form. Then a concentric spatial pooling is utilized to neutralize subjective variance while keeping variance between different gestures.
 3D Facet

3D Facet is defined as a set of local cloud points (as well as the concensus plane) locates by a center point (q as in image).

 Robust Coding

To robustly encode a 3D Facet, we project the estimated normal to three orthogonal planes,(i.e., x-y, y-z and x-z) and quantize the normal vector onto bins of each 2D plane respectively. To retain more information, we quantize each normal vector to 2 bin centers on each plane in a local-soft manner rather than simply hard assignment. In this step, we code each normal vector into a code of length (5+5+8=18).

 Concentric Spatial Pooling

In our work, we follow a dense sampling manner, namely, each pixel on the depth map is coded. To obtain a conclusive description of the hand gesture, we use a circular grid to pool all the codes to form a descriptor. The number of bins are 32 (8 angular and 4 spatial), thus the final length of H3DF descriptor is 576.
 Experimental Results

We evaluate our H3DF descriptor in two public data sets, one is from NTU's Hand Digits Recognition Data Set, which contains 10 hand gestures of digits 0~9; the other is from Surrey University, ASL Character Data Set, which contains 24 English letters from a to z while j and z are excluded since they are not static gestures. Experiments shows that our performance is much better.
 Samples of Hand Gestures
 Publications

• C. Zhang, X. Yang, and Y. Tian. Histogram of 3D Facets: A Characteristic Descriptor for Hand Gesture Recognition IEEE FG 2013

 Last Modified Jan. 2013