In the NTU hand gesture dataset, author has shared the color image as well as a depth profile of each image. 

How can I use depth profile to remove background? 

How can I convert depth profile into depth Image?

For feature extraction, can we use the depth image or should I use color images?