Abstract: To enhance the recognition accuracy of slow-moving small underwater targets using active sonar, a classification and recognition method for active sonar target echoes based on Continuous High Order Time Lacunarity (Con-HOT-Lac) and Residual Network (ResNet) is proposed. This method combines multiple frames of active sonar image sequences and processes sonar images by calculating the higher-order lacunarity of sliding sub-volumes, extracting Con-HOT-Lac inputs that can jointly utilize both beam pattern-level features and trajectory-level features. Meanwhile, ResNet is selected as the target recognition classifier, and transfer learning methods are employed to distinguish the categories of underwater slow-moving small targets based on the obtained Con-HOT-Lac inputs. The results of real-world data processing demonstrate that the proposed method has improved the average recognition accuracy for underwater slow-moving small targets, providing a feasible approach for the joint use of features at different levels.