This paper presents a natural and comfortable communication system between human and robot based on synchronization to human emotional state using human facial expression recognition. The system consists of three parts: human emotion recognition, robotic emotion generation, and robotic emotion expression. The robot recognizes human emotion through human facial expressions, and robotic emotion is generated and synchronized with human emotion dynamically using a vector field of dynamics. The robot makes dynamically varying facial expressions to express its own emotions to the human. We conducted a communication experiment to examine the effectiveness of the proposed system. We found that subjects became much more comfortable after communicating with the robot with synchronized emotions. Subjects felt somewhat uncomfortable after communicating with the robot with non-synchronized emotions. During emotional synchronization, subjects communicated much more with the robot, and the communication time was double that during non-synchronization. Furthermore, in the case of emotional synchronization, subjects had good impressions of the robot, much better than the impressions in the case of non-synchronization. It was confirmed in this study that emotional synchronization in human-robot communication can be effective in making humans comfortable, and makes the robot much more favorable and acceptable to humans.