Modified PPCA Methods for Radar HRRP Robust Recognition

Bin Li,Hui Li,Weibin Huang
DOI: https://doi.org/10.3969/j.issn.1000-2758.2016.06.015
2016-01-01
Abstract:In the condition of high signal?to?noise ratio( SNR) , utilizing probabilistic principal component analysis (PPCA) model obtained high recognition rate in radar high resolution range profile (HRRP) recognition field. But in the actual working environment, HRRPs getting from the test phase usually have low SNR which will cause mod?el mismatch problem influencing recognition rate. In order to get noise robust PPCA based methods, according to the prior mismatch information which are caused by different noise sources, the modified PPCA statistic parameters are computed to make up for the error of mismatch in model space. On the other hand, the two methods proposed in this paper directly estimate the increased noise power to avoid the step in solving optimization problem of modified parameters which greatly improve the calculation efficiency. Finally, in the simulation experiment applying the max?imum a posteriori probability classifier to test the modified PPCA model, the results show the feasibility of these two methods below 20 dB SNR.
What problem does this paper attempt to address?