In radar target recognition based on kernel method, Support vector data description(SVDD) has been applied to High resolution range profiles(HRRPs)recognition. In this paper, first, three distribution models,i.e., Membership model, Cloud model, Gaussian mixture model, are developed to describe distribution characteristics of HRRPs in extended space. Secondly, test HRRPs in multi-target hypersphere spaces are classified, in accordance with their multi-space distributing characteristics,into two types, i.e., shrink sample and slack sample. Determination of the property of a test sample is achieved by using the minimum relative distance for shrink samples and three distribution models for slack samples, respectively. Thus three HRRP recognition methods based on dual space SVDD are formed. The extensive experiment results for HRRPs of four planes show that the proposed methods have better recognition performance than the recognition method based on single space SVDD. In this paper, first, three distribution models, ie, Membership model, Cloud model, Gaussian mixture model (SVDD) has been applied to High resolution range profiles , were developed to describe distribution characteristics of HRRPs in extended space. Determination of HRRPs in multi-target hypersphere spaces are in accordance with their multi-space distribution characteristics, into two types, ie, shrink sample and slack sample. Determination of the property of a test sample is achieved by using the minimum relative distance for shrink samples and three distribution models for slack samples, respectively. respectively. The extensive experiment results for dual beds SVDD are formed. The extensive experiment results for HRRPs of four sheets show that the proposed methods have better recognition performance than the recognition method based on single space SVDD.