A proposed independence index for solving overlapped influence of multiple indicators in evaluation of HRRP recognition methods

Abstract:
In the process of using multiple indicators to evaluate recognition methods, because of the phenomenon of overlapped influence of different indicators on evaluation results, it is necessary to concern the independence of the indicators which have strong correlation with each other. A method for decreasing cross impact that mean SNR and number of the samples, as evaluation indicators, have on the quality of data set is proposed. Given frequently-used HRRP recognition methods, the independence index is defined by the area below the fitting curve obtained from the quality of data set and correct recognition rate. In the evaluation process, this new index is substituted for mean SNR and number of the samples to give a more accurate result. To verify the validity and feasibility of the proposed method, a simulated HRRP data set and four HRRP recognition algorithms were taken to make an experiment. The results show that this method effectively make the evaluation process not affected by the change of data set.