Large neural network models with high computational power and large memory have seen considerable development and application in the field of individual radiation source identification, achieving good recognition effects. Due to the limited resources of ordinary hardware devices, it is difficult to deploy complex neural networks. Therefore, this paper proposes a channel Re-Growth network pruning strategy based on the Deep Deterministic Policy Gradient (ReG-DDPG). Pruning strategies, serving as state actions in the Deep Deterministic Policy Gradient (DDPG) algorithm, enhance the efficiency of obtaining the optimal model pruning strategy through iterative updates. After pruning, the pruned channels are regrown to achieve better recognition effects. To confirm the dependability and potency of this approach, We compared the identification performance of this method with various pruning methods, demonstrating that this method is more suitable for signal recognition in resource-constrained environments.