Based on the Riemann geometric classification algorithm, we explore the possibility of decoding kinematic information of three natural reach-and-execute actions using movement-related cortical potentials ( MRCPs). EEG signals are collected from 9 healthy subjects during the execution of pinch, palmar and precision disk rotation actions that involve two levels of speeds and forces. After preprocessing, MRCPs signals are transformed into covariance space and input into minimum distance to riemannian mean (MDRM) classifier. In this way, we successfully decode the movement parameters of natural hand movements based on MRCPs. For the kinematic parameters of three hand movement, we show that the grand average result of binary classification could reach 89. 24% , and the result of multi-classification could reach 75. 28% . The riemannian framework adopted in this article is novel and efficient, which provides a new way for MRCPs classification of brain-computer interface. Meanwhile, this study is of great importance for controlling neuroprosthesis or other rehabilitation devices in a fine and natural way, which could drastically increase the acceptance of motor impaired users.