Abstract: In this paper,we introduce biased support vector machine (BSVM)-based relevance feedback into content-based sketch retrieval.The retrieval system first employs both global and structural features to represent the sketch content,and performs content matching to get the candidate results.BSVM-based learning technique is then used to implement the relevance feedback,to effectively capture users'intention and improve the retrieval performance.Experiments show that the relevance feedback based on BSVM effectively improves the retrieval performance of content-based sketch retrieval.