Abstract: Hand pose estimation is foundation of Human-Computer Interface (HCI) in virtual reality cockpit simulator but it is a challenging problem due to the variation of posture appearance, especially only from single camera. This study proposes a novel visual hand pose estimation method based on Hierarchical Temporal Memory (HTM) which is a biologically inspired model consisting of a hierarchically connected network of nodes. A database containing synthetic images generated by graphics software Pose8 and real images captured by camera is built to train the HTM network. The trained HTM network is used to classify the hand gestures and estimate the wrist parameters of input images. Subsequently, the classification result of HTM is utilized to identify hand motion sequence which is predefined and the finger parameters are acquired by searching the concrete position of input images in the sequence. Experimental results show that the proposed method possesses the characteristic of accurate rendering of the virtual hand applied in HCI and the ability to reconstruct hand postures in a virtual reality cockpit simulator.