在未知期望信号的条件下,提出一种能够抑制脉冲噪声的鲁棒递归最小二乘自适应滤波方法.与传统最小二乘法的代价函数不同,通过引入饱和非线性约束,降低可能出现的脉冲噪声对滤波器权值更新的影响.此外,提出一种多步预测器来重构滤波器的输入信号,通过比较判断滤波器输入信号可能受到脉冲噪声干扰时,采用预测值来替代原始观测信号.实验结果表明,提出的无监督鲁棒递归最小二乘自适应滤波方法在未受到脉冲噪声干扰时与传统的递归最小二乘法具有相近的收敛性能;在脉冲噪声条件下,传统递归最小二乘法和其他的无监督自适应滤波方法性能都变得很差,但本文提出的方法几乎未受到脉冲噪声的影响. In the condition of unknown signal, a robust recursive least square adaptive filtering method which can suppress impulse noise is proposed. Different from the cost function of traditional least square method, by introducing saturation nonlinear constraint, the possible impulse noise On the weight update of the filter.Furthermore, a multi-step predictor is proposed to reconstruct the input signal of the filter. When the input signal of the filter is judged to be impaired by impulse noise, the predicted value is used to replace the original observed signal. The experimental results show that the proposed unsupervised robust recursive least squares adaptive filtering method has the similar convergence performance with the traditional recursive least square method when it is not impacted by impulsive noise. Under the impulsive noise condition, the traditional recursive least square method and The performance of other unsupervised adaptive filtering methods has become very poor, but the method proposed in this paper has almost no impulse noise.