基于独立分量分析的运动声源波达方向估计

Moving Sound Source DOA Estimation Based on Independent Component Analysis

针对运动声源目标波达方向估计中,波束形成法和盲信号分离法存在不能分辨临近目标、计算量大等问题,提出采用独立分量分析算法进行估计。通过采用完全正交分解的独立分量分析实现对信号的去噪和降维,以揭示运动目标的数量,并在此基础上经GIVENS旋转实现对数据的更新和释放,采用梯度下降低通滤波选择合适的步长。坦克目标探测实验表明:此方法计算简便,收敛迅速,能准确探测目标数量,估计波达方向。

The methods of beam forming and blind signal separation on DOA estimation of moving noise objects, can not recognize nearing targets, and have a large calculation amount. Aiming at these problems, a new way of independent component analysis （ICA） was proposed in this paper. The noise and dimension reduction was car- ried out to reveal the moving targets amount by completely orthogonal ICA, based on which, the data was re- newed and released by GIVENS. Some form of gradient descent was adopted to minimize the cost function, ant the gradient could be monitored to adjust the step size appropriately. The experimental results indicated that th~ method was feasible and effective to detect moving objects.