摘要
支持向量域描述(SVDD)算法适用于异常点检测,但对于不同领域样本组的整体快速识别则力不从心.为此,基于SVDD算法提出一种基于最小包含球的领域自适应算法(MEB-DA),并将其发展为基于中心约束型最小包含球的领域自适应法(CCMEB-DA),以满足大样本数据的快速计算.该算法通过计算各自数据组的包含球球心对不同领域数据进行整体校正和相似度识别,具有较好的便捷性和自适应性.将所提出的算法应用于无限保真(WIFI)数据的室内定位和人脸识别检测,均取得了较好的效果,从而验证了所提出算法的有效性和快速性.
Support vector domain description(SVDD) is very suitable for testing a single anomaly point and is inadequate for testing the whole testing dataset.Based on SVDD,the algorithm of minimum enclosing ball for domain adaptation(MEBDA) is proposed.In order to achieve the rapid calculation for large datasets,an algorithm named center constrained minimum enclosing ball for domain adaptation(CCMEB-DA) is proposed.By calculating the center of each dataset,the dataset is corrected and the similarity of data is identified between different domains,which shows a good adaptability.The proposed method is applied to the fields of wireless fidelity(WIFI) indoor positioning and face detection,and the obtained experimental results show the effectiveness of the proposed algorithm.
出处
《控制与决策》
EI
CSCD
北大核心
2013年第2期177-182,187,共7页
Control and Decision
基金
国家自然科学基金项目(60903100
60975027)
关键词
中心约束型最小包含球
领域
最小包含球
数据校正
center constrained minimum enclosing ball(CCMEB)
domain
minimum enclosing ball
data correction