Repetition is ubiquitous in music across cultures and time periods, and the study of repeated music patterns offers insight into composition, structure, and evolution of musical works. In this paper, we propose a novel motif detection approach for polyphonic music. A local alignment-based segmentation method is designed to investigate the interrelationship between voices within a composition. For motif selection, we offer a musicologically-informed model that compares repeated patterns and evaluates their significance in the opus. Our method has been evaluated with the JKUPDD dataset, a Western classical music dataset with expert motif annotations. The evaluation results demonstrate strong performance, comparable to other state-of-the-art methods in music pattern discovery.