Knowledge tracing is a critical technology in intelligent tutoring systems, predicting students' future performance from their historical learning behaviors. While deep learning has bolstered its progress, existing models still falter in effectively modeling the dynamic shifts in students' knowledge states. This paper introduces the Dynamic Knowledge Perception for Temporal Knowledge Tracing Model (DKPKT). It combines an interaction perception module, using a dynamic attention mechanism to extract key interaction features reflecting knowledge state changes, and a temporal perception module, which dynamically updates knowledge states via knowledge-forgetting and knowledge-gain factors. Experiments demonstrate that DKPKT outperforms 11 mainstream baseline models on three real-world datasets.