Multi-stage 3D segmentation for ToF based gesture recognition system

The Human-Computer Interaction changed a lot in the last decade and it became gesture oriented. The up-to-date target is to develop smart interfaces which are free of contact. Hand is one of the favorite mean in this interaction because is a natural way in manipulating objects. This change was supported by the hardware evolution and the development of several 3D technologies. A critical step in any gesture recognition system is represented by the segmentation step. The current work proposes a new 3D segmentation method for the hand. The segmentation algorithm is based on depth information and cluster analysis. The algorithm was tested over the UPT-ToF3D-HGDB hand gesture database and it proves to be robust and fast.