Information Extraction (IE) is a critical task in natural language processing, aimed at extracting structured information from unstructured text. Traditional pipeline methods decouple Named Entity Recognition (NER) and Relation Extraction (RE) but suffer from error propagation issues. While joint models can capture task interactions, they face challenges such as task interference and high complexity. In this paper, we propose a HyperGraph Attention-based Pipeline Joint Entity-Relation Extraction method (HGAERE), which optimizes performance through two-stage independent model training. In the entity recognition stage, hypergraph attention representation learning is introduced to model text sequences as hypergraph nodes, leveraging hyperedge partitioning and span position encoding to enhance the accuracy of entity boundary and type identification. In the relation extraction stage, the context representations of entity type labels are incorporated to classify relations between entity pairs independently. Experiments demonstrate that HGAERE improves the F1 score for entity recognition by 1.2%-1.4% on datasets such as CoNLL-04 and SciERC, and the F1 score for relation extraction by 0.5%, with particularly strong performance in scenarios involving complex entity types.