<p>Improper operations in home appliance logistics have become a pressing issue for home appliance companies. However, existing intelligent recognition methods mostly use single-sensor information sources. In contrast, methods that use multiple sensor sources lack effective key improper operations information mining and cross-sensor feature aggregation learning mechanisms, resulting in suboptimal classification results for improper operations. This paper proposes a home appliance logistics improper operations recognition method based on game learning and multi-sensor information fusion to address these issues. First, improper manipulation features are automatically extracted from each sensor data using a parallel feature extraction module. Second, a game-relational sensor source discriminator is constructed with a feature extraction module, thereby refining the features of inappropriate operations and guiding their cross-sensor classification aggregation. Meanwhile, a feature difference metric loss function is introduced in the optimization objective to ensure the spatial separability of improperly manipulated features. Finally, features are fused using a recursive gate convolution module, and a classifier is used to classify inappropriate operations. The experimental results show that the method proposed in this paper can achieve a higher F1 Score of 1.38% than the advanced method while maintaining a low number of parameters and computational complexity.</p>