In recent years, there has been a rapid growth of real-time interactive video applications such as video conferencing, cloud gaming, and remote industrial control, which place higher demands on real-time performance and interactivity. Accurate evaluation of Quality of Experience (QoE) is essential for optimizing transmission algorithms in real-time interactive video. However, existing methods do not fully consider the real-time performance of video streaming and human visual perception, which affects QoE assessment. This paper proposes a blind QoE assessment method for real-time interactions. Specifically, we propose a non-intrusive QoE metrics measurement method that can obtain key information for QoE assessment without accessing the system internals or parsing encrypted traffic. Furthermore, we design a differentiated QoE assessment method for different real-time interactive scenarios, which combines end-to-end delay, playback stalling events, quality changes, and human visual perception. Experimental results show that the proposed method achieves the best performance on different video databases.