In order to solve the instability of gesture data among the current gesture control technologies, a robot gesture control system based on RFID backscatter communication is proposed in this study. Firstly, we add windows to the data for handling the discontinuity of the reflected signal in the time domain, and use the idea of relative entropy to extract the feature segmentation of dynamic gestures. Secondly, based on the Dynamic Time Warping (DTW) algorithm, we calculate the matching degree between the current segment and one-dimensional components in the training set, and we achieve gesture classification combining with the k-nearest algorithm. Finally, we complete this human-computer interaction application through the wireless Bluetooth serial port. Experimental results show that the system can perform actions of forward, backward, leftward, rightward, clockwise rotation, and stop. The correct feedback rate of the robot for gesture commands is higher than 84%, which demonstrates that the system has sound feasibility and robustness in actual deployment.