基于邻域风险最小化概率密度估计的自适应盲分离算法*

为实现由不同统计特性和概率分布平滑特性信号得到混合信号的盲分离，对基于支持向量机的邻域风险最小化概率密度估计算法进行研究，提出一种邻域函数的构造方法，将其与自然梯度批处理算法相结合，形成一种新的自适应盲分离算法；利用广义高斯模型分析了分离算法的精确度。通过仿真实验，验证了该算法能分离统计特性不同的混合信号，相比于基于经验风险最小化的方法，该方法在收敛速度和精度方面的性能有很大提高。

Adaptive blind source separation algorithm based on vicinal risk minimizing PDF estimation

The paper studied the vicinal risk minimization based PDF estimation algorithm using support vector machine, and proposed a new construction algorithm for the vicinity function. Combining with natural gradient batch algorithm, put forward a new adaptive blind source separation algorithm. Analyzed the precision of the solution farther using the generalized gauss mo-del. Carried out the several experiments, which proved that the algorithm could separate the mixtures containing signals with different statistical characteristic. Compared with the widely algorithm based on expirical risk minimization methods, the proposed algorithm has better performance both in the convergent speed and the precision.