Abstract
Spatial cognition is a fundamental component of human cognitive ability and plays an important role in navigation, object localization, route planning, spatial orientation, spatial transformation, spatial memory, and everyday problem solving. Conventional spatial cognition training methods are often constrained by low immersion, weak motivation, monotonous interaction, and insufficient objectivity in outcome assessment. To address these limitations, this study designed and validated a spatial cognition training system that integrates brain–computer interface technology with music-based game interaction. The proposed system was developed using Unity 3D and incorporated an OpenBCI-based electroencephalography (EEG) acquisition device, forming an integrated experimental platform that combines spatial cognition training, spatial cognition testing, behavioral recording, and synchronized EEG data acquisition. The system was intended to enhance user engagement during training and provide a multimodal evaluation framework based on both behavioral performance and EEG signals. In the training task, a beach-based virtual scene was created to improve ecological appeal and immersion. Spatialized 3D audio was used to simulate the position and direction of hidden target objects. Participants were required to localize the sound source based on auditory cues, navigate toward the target, identify it upon visual emergence, and complete a color-matching task between the located object and its corresponding destination item. Instead of using traditional score-based feedback, the system adopted a music performance reinforcement strategy, where successful completion of matching tasks unlocked virtual characters and instrument performance segments. This design integrated auditory guidance, visual interaction, action feedback, and motivational reinforcement. In addition, a progressive difficulty adjustment mechanism was implemented by changing the effective display distance between participants and target objects such that task demand gradually increased alongside training progress. To evaluate the effectiveness of the proposed system, a 35-day controlled experiment was conducted. A total of 32 healthy university students were recruited and randomly assigned to an experimental or control group, with 16 participants in each group. The experimental group received spatial cognition training through the proposed music-based game system, whereas the control group performed a non-spatial casual game task to control for general game exposure and training duration. Before and after the intervention, all participants completed three spatial cognition assessments: perspective taking spatial orientation test, Guilford–Zimmerman spatial orientation test, and Corsi block-tapping task. Repeated spatial cognition testing tasks were also conducted under consistent difficulty settings. During training and testing, behavioral data and EEG signals were synchronously collected. Behavioral performance was mainly measured by task completion time, while EEG analysis included coupling-based brain network characterization and support vector machine (SVM) classification based on time-domain statistical features. For the classification analysis, subject-level cross-validation was adopted to reduce the risk of data leakage caused by samples from the same participant appearing in both the training and testing sets. The results showed that the two groups were well balanced at baseline, with no significant differences in age, sex ratio, or initial spatial cognition scale scores. After Bonferroni correction, the experimental group showed a significant improvement in Guilford–Zimmerman spatial orientation test scores, whereas perspective taking spatial orientation test scores showed an improving trend and Corsi block-tapping task scores increased without reaching statistical significance. The completion time of the experimental group decreased continuously across five testing sessions, and the differences between the first and subsequent tests remained statistically significant after multiple-comparison correction. The comparison of pre–post change scores further revealed that the experimental group had greater mean improvements than the control group in all three spatial cognition measures, among which the perspective taking spatial orientation test displayed a more evident improvement trend. EEG coupling analysis indicated that the training process was accompanied by changes in inter-channel coupling patterns across multiple frequency bands. SVM classification under subject-level cross-validation suggested that EEG features showed a certain degree of separability before and after training, which may serve as auxiliary neural evidence in addition to behavioral and scale-based results, rather than direct proof of spatial cognitive improvement. Overall, the proposed system provides an integrated framework for multisensory task design, synchronized behavioral and neural data acquisition, and objective training evaluation. The findings suggest that the system has the potential to improve spatial orientation and spatial task performance in healthy young adults. Future studies should further optimize active control conditions, include larger and more diverse samples, and introduce frequency-domain, time-frequency, and feature-importance analyses to improve the interpretability and generalizability of the system. This study provides a useful reference for the development of EEG-assisted, game-based, and scientifically assessable cognitive training systems.