Physiological states and body postures can tell your flow experience——application of BP neural networks

Abstract

Accurately evaluating flow level is critical for game designers. In order to identify and analyze the real-time flow experience without any intrusion, we proposed to use body posture data to predicted flow level. The Back-propagation (BP) neural network models were developed to predict the flow experience by the physiological states and body postures during each game round. We collected 210 samples and split this data into the training and test sets by 9:1. The results showed that: (1) The body posture can effectively assess flow experience. The performance of the model training by body posture indicators is comparable to the model training by physiological indicators. (2) The prediction accuracy of the head distance is highest (i.e., 84.115%), among the body postures indexes. The prediction accuracy of high frequency heart rate variability is highest (i.e., 82.734%), among the physiological indexes. With a combination of body postures and physiological indexes, we achieve the prediction accuracy of 87.370%. The findings of this work provide a practical and effective approach to recognize the flow experience level during the mobile games.