Abstract: The existing approach of stereophonic acoustic echo cancellation is designed by a two-input/two-output system with real random variables. Its structure appears to be complex and it is difficult to realize. This scheme will be recast as a single-
input/single-output system with complex random variables by approved the widely linear model. The advantage of this method is that instead of handling two real output signals separately, it only handles one complex output signal. Moreover, the perception and quality of the stereo will be modified by the phase and module of the complex input signal, respectively. The coherence can be declined by making the input signal reasonably distorted in order to solve the nonuniqueness problem. The widely linear model and distorted signal are applied in affine projection algorithm to verify the misalignment and convergence. Experimental results indicate that the proposed method has good numerical features of less misalignment and fast convergence, so the SAEC with widely linear model has more advantage.

Key words: Stereophonic Acoustic Echo Cancellation（SAEC）, widely linear model, nonlinear distortion, Affine Projection Algorithm（APA）

摘要： 现有的立体声回声抵消器是一个实变量双输入双输出的装置，其结构复杂不易实现。宽线性模型的引入，提供了一种复变量单输入单输出的装置来替代实变量双输入双输出装置，其优点是只需处理一个复变量的输出信号而不是两个实变量输出信号，而且能通过复变量输入信号的相位和幅值分别调控声音的立体感和音质。利用输入信号适度失真的方法降低两个信号之间的相关性以解决因滤波而产生的非唯一性问题。把宽线性模型和失真信号应用到仿射投影算法中，通过仿真验证改进方法的误差性能和收敛速度。结果表明改进的方法具有误差小和收敛快的特点，因此宽线性SAEC模型更有优势。

关键词: 立体声声学回声抵消, 宽线性模型, 非线性失真, 仿射投影算法