Artificial neural networks often forget knowledge from previous tasks while adapting to new ones, a phenomenon known as catastrophic forgetting. Traditional experience replay and generative replay continual learning (CL) methods address this by reintroducing original samples or pseudo-samples during training. However, experience replay requires large memory, and generative replay has difficulty handling tasks with complex inputs. Inspired by the brain’s mechanisms of memory consolidation and replay, we propose Brain Generative Replay, a novel approach that leverages the rich semantic and detailed information encoded in electroencephalography (EEG) signals to generate high-quality, task-relevant samples. Furthermore, by uniquely incorporating human brain knowledge representation (EEG signals) as the foundation for replay, our method requires storing only minimal EEG data and dynamically generates semantically consistent yet detail-diverse replay samples by decoding multiple times, thereby improving storage efficiency. Extensive experiments demonstrate that Brain Generative Replay effectively mitigates catastrophic forgetting while balancing storage and performance.