Hand Pose Detection for Vision-based Gesture Interfaces
Luis Antón Canalís, Elena Sánchez‐Nielsen, Modesto Castrillón-Santana · Machine Vision Applications · 2005
Vision-based applications designed for humanmachine inter action r equir e fast and accur ate hand detection. However , pr evious wor k s on this field assume differ ent constr aints, lik e a limitation in the number of detected gestur es, because hands ar e highly complex objects to locate. This paper pr esents an appr oach which changes the detection tar get without limiting the number of detected gestur es. Using a cascade classifier we detect hands based on their wr ists. With this appr oach, we intr oduce two main contr ibutions: (1) a r eliable segmentation, independently of the gestur e being made and (2) a tr aining phase faster than pr evious cascade classifier based methods. The paper includes ex per imental evaluations with differ ent video str eams that illustr ate the efficiency and suitability for per ceptual inter faces.