The system is a multimodal cognitive training platform designed for real-world environments. It integrates EEG, IMU, and egocentric video data with large language models (LLMs) to understand a user's cognitive state, actions, and surroundings in real time. An IMU-assisted denoising algorithm ensures high-quality EEG signals by removing motion artifacts. The platform offers personalized training tasks that target memory, attention, and spatial awareness. This allows for effective and meaningful rehabilitation in natural, daily life settings.