Person re-identification endeavors to correlate individual identities across disjoint visual fields. In practical scenarios, individuals are frequently subject to partial occlusions, leading to missing or obscured discriminative foreground details. To address this challenge, we propose a novel module dedicated to enhancing the foreground person’s features. This module ingeniously generates a person foreground mask based on the output features of a neural network. Subsequently, this mask serves as a form of feedback, guiding the neural network to concentrate its learning efforts on extracting features around the obstructed person’s area. To further bolster the resilience of the neural network in extracting robust person features under occlusions, we introduce a noise channel discarding technique. This innovative method enhances recognition accuracy by selectively discarding channels within the network that are predominantly associated with occlusion artifacts. The efficacy of our approach is substantiated on the Occluded ReID dataset.