With the development of Human-machine interface (HMI), the requirements of perceiving the human intention are much higher. Electrical Impedance Tomography (EIT) is a promising alternative to existing HMIs because of its portability, non-invasiveness and inexpensiveness. In this study, we designed an EIT-based gesture recognition method achieving the recognition of 9 forearm motion patterns. We analysed the parameters, including current level and contact impedance, which are relevant for practical applications in robotic control. The gesture recognition method produced an average accuracy of 99.845% over nine gestures with PCA and QDA model on one subject. The preliminary results of parameter analysis suggested that the resolution increased with the current amplitude less than a threshold (5.5 mA) but decreased when the current amplitude was over 5.5 mA. The mean value of Region of Interest (ROI) nodes didn’t change obviously when the contact impedance increased. In future works, extensive studies will be conducted on the priori information of forearm and biological-model-based methods to further improve recognition performances in more complicated tasks.