基于多任务复数因子分析模型的雷达高分辨距离像识别方法

传统的高分辨距离像(HRRP)统计识别方法大部分只使用雷达目标高分辨回波的幅值信息且需要大量的训练样本保证统计模型参数学习的精度。为了充分利用高分辨回波的相位信息，在雷达采样率有限、训练样本数不足的条件下保证统计识别的性能，该文提出一种多任务学习(MTL)复数因子分析(CFA)模型，将数据描述推广到复数域，将每个方位帧训练样本的统计建模视为单一的学习任务，各学习任务共享加载矩阵，利用贝塔伯努利(Beta-Bernoulli)稀疏先验自适应地选择各任务需要的因子，完成多任务的共同学习。基于实测数据的识别实验显示，与传统的单任务学习(STL)因子分析模型相比，该文提出的多任务因子分析模型具有更低的模型复杂度且在小样本条件下可以显著提高识别性能。

Radar HRRP Target Recognition Method Based on Multi-task Learning and Complex Factor Analysis

Most traditional recognition methods for High Resolution Range Profile (HRRP) only utilize the amplitude information and need large number of training samples to obtain better estimation precision of model parameters. To utilize the phase information contained in the complex echoes and obtain better recognition performance with small training data and low sampling rate, a statistical model based on Multi-Task Leaning (MTL) and Complex Factor Analysis (CFA), referred to as MTL-CFA, is proposed in this paper. The MTL-CFA model directly describes the complex HRRP data. The statistical modeling of each training aspect-frame is considered as a single task, and all tasks share a common loading matrix. The factor number of each task is automatically determined via the Beta-Bernoulli sparse prior. Experimental results based on measured data show that the proposed model MTL-CFA can not only describe the observed data with lower order of model complexity, but also obtain satisfactory recognition accuracy with small training data, compared with the traditional Single- Task Learning (STL) based on FA models.