A novel mixed multi-index comprehensive evaluation method for HRRP recognition algorithms based on D-S evidence theory

Abstract:
Though many multi-index comprehensive evaluation methods have been used to evaluate some HRRP recognition algorithm, sometimes they couldn’t give a completely consistent result because of their own focuses and restrictions. Therefore, in order to make results more objective and recognition methods more discriminating, it is proposed to regard the results from different single evaluation methods as evidence sources to improve the quality of evaluation result. The proposed mixed method calculates the dispersion degrees and the correlation coefficients for all the individual evaluation results, measuring result discrimination and consistence, to establish a probability distribution function denoting confidence and then employs D-S evidence theory to make an integration. For verification, four evaluation methods, weighted sum method, grey correlation analysis method, TOPSIS method and COPRAS method, were selected to evaluate four HRRP recognition algorithms. The results show that not only the correlation coefficient of the proposed mixed method is higher than the four singe methods but also the dispersion degree is the highest, which proved the superiority and feasibility of this method.