Development of an EEG-Based Method for Detecting Flow State Using a Wearable Headband in a Game Environment.

Analyzing 29 EEG recordings from participants playing Tetris, a method using three entropy-based features (Slope, Distribution, and Spectral Entropy) extracted via Discrete Wavelet Transform and classified with a Random Forest model achieved 93% accuracy with random sampling and 82% with leave-one-subject-out cross-validation. The findings suggest this low-channel, consumer-device approach is promising for detecting the flow state in real-life settings.

Study at a glance

Abstract

Flow is a mental state of deep focus and immersion, associated with enhanced productivity, creativity, and well-being. While its benefits are well-documented, research on its neural basis remains emerging. EEG-based systems offer a practical approach to flow detection due to their simplicity and accessibility. Recent advancements in commercial EEG devices provide new opportunities for real-life flow detection, yet their potential remains underexplored. In this study, we analyzed 29 EEG recordings of participants playing a Tetris video game, utilizing data from a consumer-oriented EEG device to develop a low-channel method for detecting the flow state. After denoising the EEG signals, We applied the Discrete Wavelet Transform (DWT) to decompose the signals into sub-bands. From each sub-band, we extracted three entropy-based features: Slope Entropy, Distribution Entropy, and Spectral Entropy. The extracted features were subsequently fed into a Random Forest classifier using two cross-validation strategies: Random Sampling (RS) and Leave-One-Subject-Out (LOSO). The classifier demonstrated high accuracy, achieving a mean accuracy of 93% with Random Sampling validation. Additionally, the LOSO validation strategy yielded an 82% average accuracy across the dataset. These findings suggest that the proposed method is a promising approach for detecting the flow state in real-life applications.