To solve the problems of low efficiency and insufficient priority guarantee in semantic communication system resource scheduling in the scenario of multi task and multi priority flow coexistence in dynamic network environment， a deep Q-network （DQN） was used to dynamically allocate resources based on semantic priority and channel quality， and a semantic aware dynamic resource scheduling framework was proposed based on deep reinforcement learning . By introducing experience replay and target network technology to optimize the learning process， the stability and convergence efficiency of the model were enhanced. The results show that the proposed method significantly outperforms the comparison scheme in terms of bandwidth utilization （96%）， transmission delay （8.98 ms）， and high priority flow assurance （key frame accuracy of 98.3%）. Although the semantic recovery quality （2.73） is slightly lower than fixed allocation， it still meets real-time communication requirements. The proposed method effectively balances resource utilization and critical flow service quality through the dynamic optimization mechanism of semantic priority and DQN， providing efficient resource scheduling solutions for complex semantic communication systems.